7 years agomm, oom: base root bonus on current usage
David Rientjes [Thu, 30 Jan 2014 23:46:11 +0000 (15:46 -0800)]
mm, oom: base root bonus on current usage

commit 778c14affaf94a9e4953179d3e13a544ccce7707 upstream.

A 3% of system memory bonus is sometimes too excessive in comparison to
other processes.

With commit a63d83f427fb ("oom: badness heuristic rewrite"), the OOM
killer tries to avoid killing privileged tasks by subtracting 3% of
overall memory (system or cgroup) from their per-task consumption.  But
as a result, all root tasks that consume less than 3% of overall memory
are considered equal, and so it only takes 33+ privileged tasks pushing
the system out of memory for the OOM killer to do something stupid and
kill dhclient or other root-owned processes.  For example, on a 32G
machine it can't tell the difference between the 1M agetty and the 10G
fork bomb member.

The changelog describes this 3% boost as the equivalent to the global
overcommit limit being 3% higher for privileged tasks, but this is not
the same as discounting 3% of overall memory from _every privileged task
individually_ during OOM selection.

Replace the 3% of system memory bonus with a 3% of current memory usage

By giving root tasks a bonus that is proportional to their actual size,
they remain comparable even when relatively small.  In the example
above, the OOM killer will discount the 1M agetty's 256 badness points
down to 179, and the 10G fork bomb's 262144 points down to 183500 points
and make the right choice, instead of discounting both to 0 and killing
agetty because it's first in the task list.

Signed-off-by: David Rientjes <>
Reported-by: Johannes Weiner <>
Acked-by: Johannes Weiner <>
Cc: Michal Hocko <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
[bwh: Backported to 3.2: existing code changes 'points' directly rather
 than using 'adj' variable]
Signed-off-by: Ben Hutchings <>
7 years agocrypto: s390 - fix des and des3_ede ctr concurrency issue
Harald Freudenberger [Wed, 22 Jan 2014 12:01:33 +0000 (13:01 +0100)]
crypto: s390 - fix des and des3_ede ctr concurrency issue

commit ee97dc7db4cbda33e4241c2d85b42d1835bc8a35 upstream.

In s390 des and 3des ctr mode there is one preallocated page
used to speed up the en/decryption. This page is not protected
against concurrent usage and thus there is a potential of data
corruption with multiple threads.

The fix introduces locking/unlocking the ctr page and a slower
fallback solution at concurrency situations.

Signed-off-by: Harald Freudenberger <>
Signed-off-by: Herbert Xu <>
Signed-off-by: Ben Hutchings <>
7 years agocrypto: s390 - fix des and des3_ede cbc concurrency issue
Harald Freudenberger [Wed, 22 Jan 2014 12:00:04 +0000 (13:00 +0100)]
crypto: s390 - fix des and des3_ede cbc concurrency issue

commit adc3fcf1552b6e406d172fd9690bbd1395053d13 upstream.

In s390 des and des3_ede cbc mode the iv value is not protected
against concurrency access and modifications from another running
en/decrypt operation which is using the very same tfm struct
instance. This fix copies the iv to the local stack before
the crypto operation and stores the value back when done.

Signed-off-by: Harald Freudenberger <>
Signed-off-by: Herbert Xu <>
Signed-off-by: Ben Hutchings <>
7 years agocrypto: s390 - fix concurrency issue in aes-ctr mode
Harald Freudenberger [Thu, 16 Jan 2014 15:01:11 +0000 (16:01 +0100)]
crypto: s390 - fix concurrency issue in aes-ctr mode

commit 0519e9ad89e5cd6e6b08398f57c6a71d9580564c upstream.

The aes-ctr mode uses one preallocated page without any concurrency
protection. When multiple threads run aes-ctr encryption or decryption
this can lead to data corruption.

The patch introduces locking for the page and a fallback solution with
slower en/decryption performance in concurrency situations.

Signed-off-by: Harald Freudenberger <>
Signed-off-by: Herbert Xu <>
Signed-off-by: Ben Hutchings <>
7 years agos390/crypto: Don't panic after crypto instruction failures
Jan Glauber [Fri, 26 Oct 2012 13:06:12 +0000 (15:06 +0200)]
s390/crypto: Don't panic after crypto instruction failures

commit 36eb2caa7bace31b7868a57f77cb148e58d1c9f9 upstream.

Remove the BUG_ON's that check for failure or incomplete
results of the s390 hardware crypto instructions.
Rather report the errors as -EIO to the crypto layer.

Signed-off-by: Jan Glauber <>
Signed-off-by: Martin Schwidefsky <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
7 years agotarget/iscsi: Fix network portal creation race
Andy Grover [Sat, 25 Jan 2014 00:18:54 +0000 (16:18 -0800)]
target/iscsi: Fix network portal creation race

commit ee291e63293146db64668e8d65eb35c97e8324f4 upstream.

When creating network portals rapidly, such as when restoring a
configuration, LIO's code to reuse existing portals can return a false
negative if the thread hasn't run yet and set np_thread_state to
ISCSI_NP_THREAD_ACTIVE. This causes an error in the network stack
when attempting to bind to the same address/port.

This patch sets NP_THREAD_ACTIVE before the np is placed on g_np_list,
so even if the thread hasn't run yet, iscsit_get_np will return the
existing np.

Also, convert np_lock -> np_mutex + hold across adding new net portal
to g_np_list to prevent a race where two threads may attempt to create
the same network portal, resulting in one of them failing.

(nab: Add missing mutex_unlocks in iscsit_add_np failure paths)
(DanC: Fix incorrect spin_unlock -> spin_unlock_bh)

Signed-off-by: Andy Grover <>
Signed-off-by: Nicholas Bellinger <>
Signed-off-by: Ben Hutchings <>
7 years agoKVM: return an error code in kvm_vm_ioctl_register_coalesced_mmio()
Dan Carpenter [Wed, 29 Jan 2014 13:16:39 +0000 (16:16 +0300)]
KVM: return an error code in kvm_vm_ioctl_register_coalesced_mmio()

commit aac5c4226e7136c331ed384c25d5560204da10a0 upstream.

If kvm_io_bus_register_dev() fails then it returns success but it should
return an error code.

I also did a little cleanup like removing an impossible NULL test.

Fixes: 2b3c246a682c ('KVM: Make coalesced mmio use a device per zone')
Signed-off-by: Dan Carpenter <>
Signed-off-by: Paolo Bonzini <>
Signed-off-by: Ben Hutchings <>
7 years agodrm/radeon/DCE4+: clear bios scratch dpms bit (v2)
Alex Deucher [Mon, 27 Jan 2014 23:29:35 +0000 (18:29 -0500)]
drm/radeon/DCE4+: clear bios scratch dpms bit (v2)

commit 6802d4bad83f50081b2788698570218aaff8d10e upstream.

The BlankCrtc table in some DCE8 boards has some
logic shortcuts for the vbios when this bit is set.
Clear it for driver use.

v2: fix typo


Signed-off-by: Alex Deucher <>
Signed-off-by: Ben Hutchings <>
7 years agoBtrfs: setup inode location during btrfs_init_inode_locked
Chris Mason [Fri, 10 Jan 2014 01:28:00 +0000 (17:28 -0800)]
Btrfs: setup inode location during btrfs_init_inode_locked

commit 90d3e592e99b8e374ead2b45148abf506493a959 upstream.

We have a race during inode init because the BTRFS_I(inode)->location is setup
after the inode hash table lock is dropped.  btrfs_find_actor uses the location
field, so our search might not find an existing inode in the hash table if we
race with the inode init code.

This commit changes things to setup the location field sooner.  Also the find actor now
uses only the location objectid to match inodes.  For inode hashing, we just
need a unique and stable test, it doesn't have to reflect the inode numbers we
show to userland.

Signed-off-by: Chris Mason <>
[bwh: Backported to 3.2:
 - No hashval in btrfs_iget_locked()]
Signed-off-by: Ben Hutchings <>
7 years agopowerpc: Make sure "cache" directory is removed when offlining cpu
Paul Mackerras [Sat, 18 Jan 2014 10:14:47 +0000 (21:14 +1100)]
powerpc: Make sure "cache" directory is removed when offlining cpu

commit 91b973f90c1220d71923e7efe1e61f5329806380 upstream.

The code in remove_cache_dir() is supposed to remove the "cache"
subdirectory from the sysfs directory for a CPU when that CPU is
being offlined.  It tries to do this by calling kobject_put() on
the kobject for the subdirectory.  However, the subdirectory only
gets removed once the last reference goes away, and the reference
being put here may well not be the last reference.  That means
that the "cache" subdirectory may still exist when the offlining
operation has finished.  If the same CPU subsequently gets onlined,
the code tries to add a new "cache" subdirectory.  If the old
subdirectory has not yet been removed, we get a WARN_ON in the
sysfs code, with stack trace, and an error message printed on the
console.  Further, we ultimately end up with an online cpu with no
"cache" subdirectory.

This fixes it by doing an explicit kobject_del() at the point where
we want the subdirectory to go away.  kobject_del() removes the sysfs
directory even though the object still exists in memory.  The object
will get freed at some point in the future.  A subsequent onlining
operation can create a new sysfs directory, even if the old object
still exists in memory, without causing any problems.

Signed-off-by: Paul Mackerras <>
Signed-off-by: Benjamin Herrenschmidt <>
Signed-off-by: Ben Hutchings <>
7 years agoACPI / init: Flag use of ACPI and ACPI idioms for power supplies to regulator API
Mark Brown [Mon, 27 Jan 2014 00:32:14 +0000 (00:32 +0000)]
ACPI / init: Flag use of ACPI and ACPI idioms for power supplies to regulator API

commit 49a12877d2777cadcb838981c3c4f5a424aef310 upstream.

There is currently no facility in ACPI to express the hookup of voltage
regulators, the expectation is that the regulators that exist in the
system will be handled transparently by firmware if they need software
control at all. This means that if for some reason the regulator API is
enabled on such a system it should assume that any supplies that devices
need are provided by the system at all relevant times without any software

Tell the regulator core to make this assumption by calling
regulator_has_full_constraints(). Do this as soon as we know we are using
ACPI so that the information is available to the regulator core as early
as possible. This will cause the regulator core to pretend that there is
an always on regulator supplying any supply that is requested but that has
not otherwise been mapped which is the behaviour expected on a system with

Should the ability to specify regulators be added in future revisions of
ACPI then once we have support for ACPI mappings in the kernel the same
assumptions will apply. It is also likely that systems will default to a
mode of operation which does not require any interpretation of these
mappings in order to be compatible with existing operating system releases
so it should remain safe to make these assumptions even if the mappings
exist but are not supported by the kernel.

Signed-off-by: Mark Brown <>
Signed-off-by: Rafael J. Wysocki <>
Signed-off-by: Ben Hutchings <>
7 years agobtrfs: restrict snapshotting to own subvolumes
David Sterba [Wed, 15 Jan 2014 17:15:52 +0000 (18:15 +0100)]
btrfs: restrict snapshotting to own subvolumes

commit d024206133ce21936b3d5780359afc00247655b7 upstream.

Currently, any user can snapshot any subvolume if the path is accessible and
thus indirectly create and keep files he does not own under his direcotries.
This is not possible with traditional directories.

In security context, a user can snapshot root filesystem and pin any
potentially buggy binaries, even if the updates are applied.

All the snapshots are visible to the administrator, so it's possible to
verify if there are suspicious snapshots.

Another more practical problem is that any user can pin the space used
by eg. root and cause ENOSPC.

Original report:

Signed-off-by: David Sterba <>
Signed-off-by: Josef Bacik <>
Signed-off-by: Chris Mason <>
[bwh: Backported to 3.2:
 - Adjust context
 - Use the same cleanup code for success and error cases, as done
   upstream in commit ecd188159efa
   ('switch btrfs_ioctl_snap_create_transid() to fget_light()')]
Signed-off-by: Ben Hutchings <>
7 years agoBtrfs: handle EAGAIN case properly in btrfs_drop_snapshot()
Wang Shilong [Tue, 7 Jan 2014 09:26:58 +0000 (17:26 +0800)]
Btrfs: handle EAGAIN case properly in btrfs_drop_snapshot()

commit 90515e7f5d7d24cbb2a4038a3f1b5cfa2921aa17 upstream.

We may return early in btrfs_drop_snapshot(), we shouldn't
call btrfs_std_err() for this case, fix it.

Signed-off-by: Wang Shilong <>
Signed-off-by: Josef Bacik <>
Signed-off-by: Chris Mason <>
Signed-off-by: Ben Hutchings <>
7 years agoKVM: PPC: e500: Fix bad address type in deliver_tlb_misss()
Mihai Caraman [Thu, 9 Jan 2014 15:01:05 +0000 (17:01 +0200)]
KVM: PPC: e500: Fix bad address type in deliver_tlb_misss()

commit 70713fe315ed14cd1bb07d1a7f33e973d136ae3d upstream.

Use gva_t instead of unsigned int for eaddr in deliver_tlb_miss().

Signed-off-by: Mihai Caraman <>
Signed-off-by: Alexander Graf <>
[bwh: Backported to 3.2: adjust filename]
Signed-off-by: Ben Hutchings <>
7 years agoi2c: piix4: Add support for AMD ML and CZ SMBus changes
Shane Huang [Wed, 22 Jan 2014 22:05:46 +0000 (14:05 -0800)]
i2c: piix4: Add support for AMD ML and CZ SMBus changes

commit 032f708bc4f6da868ec49dac48ddf3670d8035d3 upstream.

The locations of SMBus register base address and enablement bit are changed
from AMD ML, which need this patch to be supported.

Signed-off-by: Shane Huang <>
Reviewed-by: Jean Delvare <>
Signed-off-by: Wolfram Sang <>
[bwh: Backported to 3.2:
 - Adjust context
 - Aux bus support is not included]
Signed-off-by: Ben Hutchings <>
7 years agob43: fix the wrong assignment of status.freq in b43_rx()
ZHAO Gang [Fri, 17 Jan 2014 16:17:38 +0000 (00:17 +0800)]
b43: fix the wrong assignment of status.freq in b43_rx()

commit 64e5acb09ca6b50c97299cff9ef51299470b29f2 upstream.

Use the right function to update frequency value.

If rx skb is probe response or beacon, the wrong frequency value can
cause problem that bss info can't be updated when it should be.

Fixes: 8318d78a44d4 ("cfg80211 API for channels/bitrates, mac80211
and driver conversion")
Signed-off-by: ZHAO Gang <>
Acked-by: Larry Finger <>
Signed-off-by: John W. Linville <>
Signed-off-by: Ben Hutchings <>
7 years agoore: Fix wrong math in allocation of per device BIO
Boaz Harrosh [Thu, 21 Nov 2013 15:58:08 +0000 (17:58 +0200)]
ore: Fix wrong math in allocation of per device BIO

commit aad560b7f63b495f48a7232fd086c5913a676e6f upstream.

At IO preparation we calculate the max pages at each device and
allocate a BIO per device of that size. The calculation was wrong
on some unaligned corner cases offset/length combination and would
make prepare return with -ENOMEM. This would be bad for pnfs-objects
that would in that case IO through MDS. And fatal for exofs were it
would fail writes with EIO.

Fix it by doing the proper math, that will work in all cases. (I
ran a test with all possible offset/length combinations this time

Also when reading we do not need to allocate for the parity units
since we jump over them.

Also lower the max_io_length to take into account the parity pages
so not to allocate BIOs bigger than PAGE_SIZE

Signed-off-by: Boaz Harrosh <>
Signed-off-by: Ben Hutchings <>
7 years agoIB/qib: Fix QP check when looping back to/from QP1
Ira Weiny [Wed, 18 Dec 2013 16:41:37 +0000 (08:41 -0800)]
IB/qib: Fix QP check when looping back to/from QP1

commit 6e0ea9e6cbcead7fa8c76e3e3b9de4a50c5131c5 upstream.

The GSI QP type is compatible with and should be allowed to send data
to/from any UD QP.  This was found when testing ibacm on the same node
as an SA.

Reviewed-by: Mike Marciniszyn <>
Signed-off-by: Ira Weiny <>
Signed-off-by: Roland Dreier <>
Signed-off-by: Ben Hutchings <>
7 years agofuse: fix pipe_buf_operations
Miklos Szeredi [Wed, 22 Jan 2014 18:36:57 +0000 (19:36 +0100)]
fuse: fix pipe_buf_operations

commit 28a625cbc2a14f17b83e47ef907b2658576a32aa upstream.

Having this struct in module memory could Oops when if the module is
unloaded while the buffer still persists in a pipe.

Since sock_pipe_buf_ops is essentially the same as fuse_dev_pipe_buf_steal
merge them into nosteal_pipe_buf_ops (this is the same as
default_pipe_buf_ops except stealing the page from the buffer is not

Reported-by: Al Viro <>
Signed-off-by: Miklos Szeredi <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
7 years agointel-iommu: fix off-by-one in pagetable freeing
Alex Williamson [Tue, 21 Jan 2014 23:48:18 +0000 (15:48 -0800)]
intel-iommu: fix off-by-one in pagetable freeing

commit 08336fd218e087cc4fcc458e6b6dcafe8702b098 upstream.

dma_pte_free_level() has an off-by-one error when checking whether a pte
is completely covered by a range.  Take for example the case of
attempting to free pfn 0x0 - 0x1ff, ie.  512 entries covering the first
2M superpage.

The level_size() is 0x200 and we test:

  static void dma_pte_free_level(...

if (!(0 > 0 || 0x1ff < 0 + 0x200)) {

Clearly the 2nd test is true, which means we fail to take the branch to
clear and free the pagetable entry.  As a result, we're leaking
pagetables and failing to install new pages over the range.

This was found with a PCI device assigned to a QEMU guest using vfio-pci
without a VGA device present.  The first 1M of guest address space is
mapped with various combinations of 4K pages, but eventually the range
is entirely freed and replaced with a 2M contiguous mapping.
intel-iommu errors out with something like:

  ERROR: DMA PTE for vPFN 0x0 already set (to 5c2b8003 not 849c00083)

In this case 5c2b8003 is the pointer to the previous leaf page that was
neither freed nor cleared and 849c00083 is the superpage entry that
we're trying to replace it with.

Signed-off-by: Alex Williamson <>
Cc: David Woodhouse <>
Cc: Joerg Roedel <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
7 years agohp_accel: Add a new PnP ID HPQ6007 for new HP laptops
Takashi Iwai [Mon, 13 Jan 2014 11:32:44 +0000 (12:32 +0100)]
hp_accel: Add a new PnP ID HPQ6007 for new HP laptops

commit b0ad4ff35d479a46a3b995a299db9aeb097acfce upstream.

The DriveGuard chips on the new HP laptops are with a new PnP ID
"HPQ6007".  It should be compatible with older chips.

Acked-by: Éric Piel <>
Signed-off-by: Takashi Iwai <>
Signed-off-by: Matthew Garrett <>
Signed-off-by: Ben Hutchings <>
7 years agodrm/radeon: set the full cache bit for fences on r7xx+
Alex Deucher [Thu, 16 Jan 2014 23:11:47 +0000 (18:11 -0500)]
drm/radeon: set the full cache bit for fences on r7xx+

commit d45b964a22cad962d3ede1eba8d24f5cee7b2a92 upstream.

Needed to properly flush the read caches for fences.

Signed-off-by: Alex Deucher <>
[bwh: Backported to 3.2:
 - Adjust context
 - s/\bring\b/rdev/]
Signed-off-by: Ben Hutchings <>
7 years agodrm/radeon: disable ss on DP for DCE3.x
Alex Deucher [Mon, 13 Jan 2014 21:47:05 +0000 (16:47 -0500)]
drm/radeon: disable ss on DP for DCE3.x

commit d8e24525094200601236fa64a54cf73e3d682f2e upstream.

Seems to cause problems with certain DP monitors.


Signed-off-by: Alex Deucher <>
[bwh: Backported to 3.2: s/radeon_crtc->//]
Signed-off-by: Ben Hutchings <>
7 years agoturbostat: Use GCC's CPUID functions to support PIC
Josh Triplett [Wed, 21 Aug 2013 00:20:14 +0000 (17:20 -0700)]
turbostat: Use GCC's CPUID functions to support PIC

commit 2b92865e648ce04a39fda4f903784a5d01ecb0dc upstream.

turbostat uses inline assembly to call cpuid.  On 32-bit x86, on systems
that have certain security features enabled by default that make -fPIC
the default, this causes a build error:

turbostat.c: In function ‘check_cpuid’:
turbostat.c:1906:2: error: PIC register clobbered by ‘ebx’ in ‘asm’
  asm("cpuid" : "=a" (fms), "=c" (ecx), "=d" (edx) : "a" (1) : "ebx");

GCC provides a header cpuid.h, containing a __get_cpuid function that
works with both PIC and non-PIC.  (On PIC, it saves and restores ebx
around the cpuid instruction.)  Use that instead.

Signed-off-by: Josh Triplett <>
Signed-off-by: Len Brown <>
Signed-off-by: Ben Hutchings <>
7 years agommc: sdhci: fix lockdep error in tuning routine
Aisheng Dong [Mon, 23 Dec 2013 11:13:04 +0000 (19:13 +0800)]
mmc: sdhci: fix lockdep error in tuning routine

commit 2b35bd83467df6f8284b9148d6f768148c3a5e5f upstream.

The sdhci_execute_tuning routine gets lock separately by
It will cause the following lockdep error message since the &host->lock
could also be got in irq context.
Use spin_lock_irqsave/spin_unlock_restore instead to get rid of
this error message.

[ INFO: inconsistent lock state ]
3.13.0-rc1+ #287 Not tainted
inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
kworker/u2:1/33 [HC0[0]:SC0[0]:HE1:SE1] takes:
 (&(&host->lock)->rlock){?.-...}, at: [<8045f7f4>] sdhci_execute_tuning+0x4c/0x710
{IN-HARDIRQ-W} state was registered at:
  [<8005f030>] mark_lock+0x140/0x6ac
  [<80060760>] __lock_acquire+0xb30/0x1cbc
  [<800620d0>] lock_acquire+0x70/0x84
  [<8061d1c8>] _raw_spin_lock+0x30/0x40
  [<804605cc>] sdhci_irq+0x24/0xa68
  [<8006b1d4>] handle_irq_event_percpu+0x54/0x18c
  [<8006b350>] handle_irq_event+0x44/0x64
  [<8006e50c>] handle_fasteoi_irq+0xa0/0x170
  [<8006a8f0>] generic_handle_irq+0x30/0x44
  [<8000f238>] handle_IRQ+0x54/0xbc
  [<8000864c>] gic_handle_irq+0x30/0x64
  [<80013024>] __irq_svc+0x44/0x5c
  [<80329bf4>] dev_vprintk_emit+0x50/0x58
  [<80329c24>] dev_printk_emit+0x28/0x30
  [<80329fec>] __dev_printk+0x4c/0x90
  [<8032a180>] dev_err+0x3c/0x48
  [<802dd4f0>] _regulator_get+0x158/0x1cc
  [<802dd5b4>] regulator_get_optional+0x18/0x1c
  [<80461df4>] sdhci_add_host+0x42c/0xbd8
  [<80464820>] sdhci_esdhc_imx_probe+0x378/0x67c
  [<8032ee88>] platform_drv_probe+0x20/0x50
  [<8032d48c>] driver_probe_device+0x118/0x234
  [<8032d690>] __driver_attach+0x9c/0xa0
  [<8032b89c>] bus_for_each_dev+0x68/0x9c
  [<8032cf44>] driver_attach+0x20/0x28
  [<8032cbc8>] bus_add_driver+0x148/0x1f4
  [<8032dce0>] driver_register+0x80/0x100
  [<8032ee54>] __platform_driver_register+0x50/0x64
  [<8084b094>] sdhci_esdhc_imx_driver_init+0x18/0x20
  [<80008980>] do_one_initcall+0x108/0x16c
  [<8081cca4>] kernel_init_freeable+0x10c/0x1d0
  [<80611b28>] kernel_init+0x10/0x120
  [<8000e9c8>] ret_from_fork+0x14/0x2c
irq event stamp: 805
hardirqs last  enabled at (805): [<8061d43c>] _raw_spin_unlock_irqrestore+0x38/0x4c
hardirqs last disabled at (804): [<8061d2c8>] _raw_spin_lock_irqsave+0x24/0x54
softirqs last  enabled at (570): [<8002b824>] __do_softirq+0x1c4/0x290
softirqs last disabled at (561): [<8002bcf4>] irq_exit+0xb4/0x10c

other info that might help us debug this:
 Possible unsafe locking scenario:


 *** DEADLOCK ***

2 locks held by kworker/u2:1/33:
 #0:  (kmmcd){.+.+..}, at: [<8003db18>] process_one_work+0x128/0x468
 #1:  ((&(&host->detect)->work)){+.+...}, at: [<8003db18>] process_one_work+0x128/0x468

stack backtrace:
CPU: 0 PID: 33 Comm: kworker/u2:1 Not tainted 3.13.0-rc1+ #287
Workqueue: kmmcd mmc_rescan
[<80012160>] (dump_backtrace+0x0/0x10c) from [<80012438>] (show_stack+0x18/0x1c)
 r6:bfad0900 r5:00000000 r4:8088ecc8 r3:bfad0900
[<80012420>] (show_stack+0x0/0x1c) from [<806169ec>] (dump_stack+0x84/0x9c)
[<80616968>] (dump_stack+0x0/0x9c) from [<806147b4>] (print_usage_bug+0x260/0x2d0)
 r5:8076ba88 r4:80977410
[<80614554>] (print_usage_bug+0x0/0x2d0) from [<8005f0d0>] (mark_lock+0x1e0/0x6ac)
 r9:8005e678 r8:00000000 r7:bfad0900 r6:00001015 r5:bfad0cd0
[<8005eef0>] (mark_lock+0x0/0x6ac) from [<80060234>] (__lock_acquire+0x604/0x1cbc)
[<8005fc30>] (__lock_acquire+0x0/0x1cbc) from [<800620d0>] (lock_acquire+0x70/0x84)
[<80062060>] (lock_acquire+0x0/0x84) from [<8061d1c8>] (_raw_spin_lock+0x30/0x40)
 r7:00000000 r6:bfb63000 r5:00000000 r4:bfb60568
[<8061d198>] (_raw_spin_lock+0x0/0x40) from [<8045f7f4>] (sdhci_execute_tuning+0x4c/0x710)
[<8045f7a8>] (sdhci_execute_tuning+0x0/0x710) from [<80453454>] (mmc_sd_init_card+0x5f8/0x660)
[<80452e5c>] (mmc_sd_init_card+0x0/0x660) from [<80453748>] (mmc_attach_sd+0xb4/0x180)
 r9:bf92d400 r8:8065f364 r7:00061a80 r6:bfb60000 r5:8065f358
[<80453694>] (mmc_attach_sd+0x0/0x180) from [<8044d9f8>] (mmc_rescan+0x284/0x2f0)
 r5:8065f358 r4:bfb602f8
[<8044d774>] (mmc_rescan+0x0/0x2f0) from [<8003db94>] (process_one_work+0x1a4/0x468)
 r8:00000000 r7:bfb55eb0 r6:bf80dc00 r5:bfb602f8 r4:bfb35980
[<8003d9f0>] (process_one_work+0x0/0x468) from [<8003e850>] (worker_thread+0x118/0x3e0)
[<8003e738>] (worker_thread+0x0/0x3e0) from [<80044de0>] (kthread+0xd4/0xf0)
[<80044d0c>] (kthread+0x0/0xf0) from [<8000e9c8>] (ret_from_fork+0x14/0x2c)
 r7:00000000 r6:00000000 r5:80044d0c r4:bfb37b40

Signed-off-by: Dong Aisheng <>
Signed-off-by: Chris Ball <>
[bwh: Backported to 3.2:
 - Adjust context
 - There's no platform_execute_tuning hook]
Signed-off-by: Ben Hutchings <>
7 years agolibata: disable LPM for some WD SATA-I devices
Tejun Heo [Thu, 16 Jan 2014 14:47:17 +0000 (09:47 -0500)]
libata: disable LPM for some WD SATA-I devices

commit ecd75ad514d73efc1bbcc5f10a13566c3ace5f53 upstream.

For some reason, some early WD drives spin up and down drives
erratically when the link is put into slumber mode which can reduce
the life expectancy of the device significantly.  Unfortunately, we
don't have full list of devices and given the nature of the issue it'd
be better to err on the side of false positives than the other way
around.  Let's disable LPM on all WD devices which match one of the
known problematic model prefixes and are SATA-I.

As horkage list doesn't support matching SATA capabilities, this is
implemented as two horkages - WD_BROKEN_LPM and NOLPM.  The former is
set for the known prefixes and sets the latter if the matched device
is SATA-I.

Note that this isn't optimal as this disables all LPM operations and
partial link power state reportedly works fine on these; however, the
way LPM is implemented in libata makes it difficult to precisely map
libata LPM setting to specific link power state.  Well, these devices
are already fairly outdated.  Let's just disable whole LPM for now.

Signed-off-by: Tejun Heo <>
Reported-and-tested-by: Nikos Barkas <>
Reported-and-tested-by: Ioannis Barkas <>
[bwh: Backported to 3.2:
 - Adjust context
 - Use literal 76 instead of ATA_ID_SATA_CAPABILITY]
Signed-off-by: Ben Hutchings <>
7 years agox86: Add check for number of available vectors before CPU down
Prarit Bhargava [Mon, 13 Jan 2014 11:51:01 +0000 (06:51 -0500)]
x86: Add check for number of available vectors before CPU down

commit da6139e49c7cb0f4251265cb5243b8d220adb48d upstream.


When a cpu is downed on a system, the irqs on the cpu are assigned to
other cpus.  It is possible, however, that when a cpu is downed there
aren't enough free vectors on the remaining cpus to account for the
vectors from the cpu that is being downed.

This results in an interesting "overflow" condition where irqs are
"assigned" to a CPU but are not handled.

For example, when downing cpus on a 1-64 logical processor system:

[  232.021745] smpboot: CPU 61 is now offline
[  238.480275] smpboot: CPU 62 is now offline
[  245.991080] ------------[ cut here ]------------
[  245.996270] WARNING: CPU: 0 PID: 0 at net/sched/sch_generic.c:264 dev_watchdog+0x246/0x250()
[  246.005688] NETDEV WATCHDOG: p786p1 (ixgbe): transmit queue 0 timed out
[  246.013070] Modules linked in: lockd sunrpc iTCO_wdt iTCO_vendor_support sb_edac ixgbe microcode e1000e pcspkr joydev edac_core lpc_ich ioatdma ptp mdio mfd_core i2c_i801 dca pps_core i2c_core wmi acpi_cpufreq isci libsas scsi_transport_sas
[  246.037633] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 3.12.0+ #14
[  246.044451] Hardware name: Intel Corporation S4600LH ........../SVRBD-ROW_T, BIOS SE5C600.86B.01.08.0003.022620131521 02/26/2013
[  246.057371]  0000000000000009 ffff88081fa03d40 ffffffff8164fbf6 ffff88081fa0ee48
[  246.065728]  ffff88081fa03d90 ffff88081fa03d80 ffffffff81054ecc ffff88081fa13040
[  246.074073]  0000000000000000 ffff88200cce0000 0000000000000040 0000000000000000
[  246.082430] Call Trace:
[  246.085174]  <IRQ>  [<ffffffff8164fbf6>] dump_stack+0x46/0x58
[  246.091633]  [<ffffffff81054ecc>] warn_slowpath_common+0x8c/0xc0
[  246.098352]  [<ffffffff81054fb6>] warn_slowpath_fmt+0x46/0x50
[  246.104786]  [<ffffffff815710d6>] dev_watchdog+0x246/0x250
[  246.110923]  [<ffffffff81570e90>] ? dev_deactivate_queue.constprop.31+0x80/0x80
[  246.119097]  [<ffffffff8106092a>] call_timer_fn+0x3a/0x110
[  246.125224]  [<ffffffff8106280f>] ? update_process_times+0x6f/0x80
[  246.132137]  [<ffffffff81570e90>] ? dev_deactivate_queue.constprop.31+0x80/0x80
[  246.140308]  [<ffffffff81061db0>] run_timer_softirq+0x1f0/0x2a0
[  246.146933]  [<ffffffff81059a80>] __do_softirq+0xe0/0x220
[  246.152976]  [<ffffffff8165fedc>] call_softirq+0x1c/0x30
[  246.158920]  [<ffffffff810045f5>] do_softirq+0x55/0x90
[  246.164670]  [<ffffffff81059d35>] irq_exit+0xa5/0xb0
[  246.170227]  [<ffffffff8166062a>] smp_apic_timer_interrupt+0x4a/0x60
[  246.177324]  [<ffffffff8165f40a>] apic_timer_interrupt+0x6a/0x70
[  246.184041]  <EOI>  [<ffffffff81505a1b>] ? cpuidle_enter_state+0x5b/0xe0
[  246.191559]  [<ffffffff81505a17>] ? cpuidle_enter_state+0x57/0xe0
[  246.198374]  [<ffffffff81505b5d>] cpuidle_idle_call+0xbd/0x200
[  246.204900]  [<ffffffff8100b7ae>] arch_cpu_idle+0xe/0x30
[  246.210846]  [<ffffffff810a47b0>] cpu_startup_entry+0xd0/0x250
[  246.217371]  [<ffffffff81646b47>] rest_init+0x77/0x80
[  246.223028]  [<ffffffff81d09e8e>] start_kernel+0x3ee/0x3fb
[  246.229165]  [<ffffffff81d0989f>] ? repair_env_string+0x5e/0x5e
[  246.235787]  [<ffffffff81d095a5>] x86_64_start_reservations+0x2a/0x2c
[  246.242990]  [<ffffffff81d0969f>] x86_64_start_kernel+0xf8/0xfc
[  246.249610] ---[ end trace fb74fdef54d79039 ]---
[  246.254807] ixgbe 0000:c2:00.0 p786p1: initiating reset due to tx timeout
[  246.262489] ixgbe 0000:c2:00.0 p786p1: Reset adapter
Last login: Mon Nov 11 08:35:14 from
[root@(none) ~]# [  246.792676] ixgbe 0000:c2:00.0 p786p1: detected SFP+: 5
[  249.231598] ixgbe 0000:c2:00.0 p786p1: NIC Link is Up 10 Gbps, Flow Control: RX/TX
[  246.792676] ixgbe 0000:c2:00.0 p786p1: detected SFP+: 5
[  249.231598] ixgbe 0000:c2:00.0 p786p1: NIC Link is Up 10 Gbps, Flow Control: RX/TX

(last lines keep repeating.  ixgbe driver is dead until module reload.)

If the downed cpu has more vectors than are free on the remaining cpus on the
system, it is possible that some vectors are "orphaned" even though they are
assigned to a cpu.  In this case, since the ixgbe driver had a watchdog, the
watchdog fired and notified that something was wrong.

This patch adds a function, check_vectors(), to compare the number of vectors
on the CPU going down and compares it to the number of vectors available on
the system.  If there aren't enough vectors for the CPU to go down, an
error is returned and propogated back to userspace.

v2: Do not need to look at percpu irqs
v3: Need to check affinity to prevent counting of MSIs in IOAPIC Lowest
    Priority Mode
v4: Additional changes suggested by Gong Chen.
v5/v6/v7/v8: Updated comment text

Signed-off-by: Prarit Bhargava <>
Reviewed-by: Gong Chen <>
Cc: Andi Kleen <>
Cc: Michel Lespinasse <>
Cc: Seiji Aguchi <>
Cc: Yang Zhang <>
Cc: Paul Gortmaker <>
Cc: Janet Morgan <>
Cc: Tony Luck <>
Cc: Ruiv Wang <>
Cc: Gong Chen <>
Signed-off-by: H. Peter Anvin <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
7 years agomd/raid5: fix long-standing problem with bitmap handling on write failure.
NeilBrown [Wed, 15 Jan 2014 22:35:38 +0000 (09:35 +1100)]
md/raid5: fix long-standing problem with bitmap handling on write failure.

commit 9f97e4b128d2ea90a5f5063ea0ee3b0911f4c669 upstream.

Before a write starts we set a bit in the write-intent bitmap.
When the write completes we clear that bit if the write was successful
to all devices.  However if the write wasn't fully successful we
should not clear the bit.  If the faulty drive is subsequently
re-added, the fact that the bit is still set ensure that we will
re-write the data that is missing.

This logic is mediated by the STRIPE_DEGRADED flag - we only clear the
bitmap bit when this flag is not set.
Currently we correctly set the flag if a write starts when some
devices are failed or missing.  But we do *not* set the flag if some
device failed during the write attempt.
This is wrong and can result in clearing the bit inappropriately.

So: set the flag when a write fails.

This bug has been present since bitmaps were introduces, so the fix is
suitable for any -stable kernel.

Reported-by: Ethan Wilson <>
Signed-off-by: NeilBrown <>
[bwh: Backported to 3.2: adjust context, indentation]
Signed-off-by: Ben Hutchings <>
7 years agoKVM: x86: limit PIT timer frequency
Marcelo Tosatti [Mon, 6 Jan 2014 14:00:02 +0000 (12:00 -0200)]
KVM: x86: limit PIT timer frequency

commit 9ed96e87c5748de4c2807ef17e81287c7304186c upstream.

Limit PIT timer frequency similarly to the limit applied by
LAPIC timer.

Reviewed-by: Jan Kiszka <>
Signed-off-by: Marcelo Tosatti <>
Signed-off-by: Paolo Bonzini <>
[bwh: Backported to 3.2:
 - Adjust context
 - s/ps->period/pt->period/]
Signed-off-by: Ben Hutchings <>
7 years agodm sysfs: fix a module unload race
Mikulas Patocka [Tue, 14 Jan 2014 00:37:54 +0000 (19:37 -0500)]
dm sysfs: fix a module unload race

commit 2995fa78e423d7193f3b57835f6c1c75006a0315 upstream.

This reverts commit be35f48610 ("dm: wait until embedded kobject is
released before destroying a device") and provides an improved fix.

The kobject release code that calls the completion must be placed in a
non-module file, otherwise there is a module unload race (if the process
calling dm_kobject_release is preempted and the DM module unloaded after
the completion is triggered, but before dm_kobject_release returns).

To fix this race, this patch moves the completion code to dm-builtin.c
which is always compiled directly into the kernel if BLK_DEV_DM is

The patch introduces a new dm_kobject_holder structure, its purpose is
to keep the completion and kobject in one place, so that it can be
accessed from non-module code without the need to export the layout of
struct mapped_device to that code.

Signed-off-by: Mikulas Patocka <>
Signed-off-by: Mike Snitzer <>
[bwh: Backported to 3.2:
 - Adjust context
 - Remove paranoid check of container_of() result in dm_get_from_kobject(),
   which would now be incorrect
 - Include <linux/export.h> in dm-builtin.c, included indirectly upstream]
Signed-off-by: Ben Hutchings <>
7 years agousb: ehci: add freescale imx28 special write register method
Peter Chen [Fri, 10 Jan 2014 05:51:26 +0000 (13:51 +0800)]
usb: ehci: add freescale imx28 special write register method

commit feffe09f510c475df082546815f9e4a573f6a233 upstream.

According to Freescale imx28 Errata, "ENGR119653 USB: ARM to USB
register error issue", All USB register write operations must
use the ARM SWP instruction. So, we implement a special ehci_write
for imx28.

Discussion for it at below:

Without this patcheset, imx28 works unstable at high AHB bus loading.
If the bus loading is not high, the imx28 usb can work well at the most
of time. There is a IC errata for this problem, usually, we consider
IC errata is a problem not a new feature, and this workaround is needed
for that, so we need to add them to stable tree 3.11+.

Signed-off-by: Peter Chen <>
Acked-by: Alan Stern <>
Signed-off-by: Marc Kleine-Budde <>
Tested-by: Marc Kleine-Budde <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2:adjust context]
Signed-off-by: Ben Hutchings <>
7 years agonfs4.1: properly handle ENOTSUP in SECINFO_NO_NAME
Weston Andros Adamson [Mon, 13 Jan 2014 21:54:45 +0000 (16:54 -0500)]
nfs4.1: properly handle ENOTSUP in SECINFO_NO_NAME

commit 78b19bae0813bd6f921ca58490196abd101297bd upstream.

Don't check for -NFS4ERR_NOTSUPP, it's already been mapped to -ENOTSUPP
by nfs4_stat_to_errno.

This allows the client to mount v4.1 servers that don't support
SECINFO_NO_NAME by falling back to the "guess and check" method of

Signed-off-by: Weston Andros Adamson <>
Signed-off-by: Trond Myklebust <>
Signed-off-by: Ben Hutchings <>
7 years agortlwifi: rtl8192cu: Add new device ID
Larry Finger [Thu, 9 Jan 2014 16:27:27 +0000 (10:27 -0600)]
rtlwifi: rtl8192cu: Add new device ID

commit f87f960b2fb802f26ee3b00c19320e57a9c583ff upstream.

Reported-by: Jan Prinsloo <>
Tested-by: Jan Prinsloo <>
Signed-off-by: Larry Finger <>
Signed-off-by: John W. Linville <>
Signed-off-by: Ben Hutchings <>
7 years agoftrace: Fix synchronization location disabling and freeing ftrace_ops
Steven Rostedt (Red Hat) [Mon, 13 Jan 2014 17:56:21 +0000 (12:56 -0500)]
ftrace: Fix synchronization location disabling and freeing ftrace_ops

commit a4c35ed241129dd142be4cadb1e5a474a56d5464 upstream.

The synchronization needed after ftrace_ops are unregistered must happen
after the callback is disabled from becing called by functions.

The current location happens after the function is being removed from the
internal lists, but not after the function callbacks were disabled, leaving
the functions susceptible of being called after their callbacks are freed.

This affects perf and any externel users of function tracing (LTTng and

Fixes: cdbe61bfe704 "ftrace: Allow dynamically allocated function tracers"
Signed-off-by: Steven Rostedt <>
[bwh: Backported to 3.2: drop change for control ops]
Signed-off-by: Ben Hutchings <>
Acked-by: Steven Rostedt <>
7 years agoftrace: Use schedule_on_each_cpu() as a heavy synchronize_sched()
Steven Rostedt [Wed, 29 May 2013 00:01:16 +0000 (20:01 -0400)]
ftrace: Use schedule_on_each_cpu() as a heavy synchronize_sched()

commit 7614c3dc74733dff4b0e774f7a894b9ea6ec508c upstream.

The function tracer uses preempt_disable/enable_notrace() for
synchronization between reading registered ftrace_ops and unregistering

Most of the ftrace_ops are global permanent structures that do not
require this synchronization. That is, ops may be added and removed from
the hlist but are never freed, and wont hurt if a synchronization is

But this is not true for dynamically created ftrace_ops or control_ops,
which are used by the perf function tracing.

The problem here is that the function tracer can be used to trace
kernel/user context switches as well as going to and from idle.
Basically, it can be used to trace blind spots of the RCU subsystem.
This means that even though preempt_disable() is done, a
synchronize_sched() will ignore CPUs that haven't made it out of user
space or idle. These can include functions that are being traced just
before entering or exiting the kernel sections.

To implement the RCU synchronization, instead of using
synchronize_sched() the use of schedule_on_each_cpu() is performed. This
means that when a dynamically allocated ftrace_ops, or a control ops is
being unregistered, all CPUs must be touched and execute a ftrace_sync()
stub function via the work queues. This will rip CPUs out from idle or
in dynamic tick mode. This only happens when a user disables perf
function tracing or other dynamically allocated function tracers, but it
allows us to continue to debug RCU and context tracking with function

Cc: "Paul E. McKenney" <>
Cc: Tejun Heo <>
Cc: Ingo Molnar <>
Cc: Frederic Weisbecker <>
Cc: Jiri Olsa <>
Cc: Peter Zijlstra <>
Acked-by: Paul E. McKenney <>
Signed-off-by: Steven Rostedt <>
[bwh: Backported to 3.2: drop change for control ops]
Signed-off-by: Ben Hutchings <>
7 years agommc: atmel-mci: fix timeout errors in SDIO mode when using DMA
Ludovic Desroches [Wed, 20 Nov 2013 15:01:11 +0000 (16:01 +0100)]
mmc: atmel-mci: fix timeout errors in SDIO mode when using DMA

commit 66b512eda74d59b17eac04c4da1b38d82059e6c9 upstream.

With some SDIO devices, timeout errors can happen when reading data.
To solve this issue, the DMA transfer has to be activated before sending
the command to the device. This order is incorrect in PDC mode. So we
have to take care if we are using DMA or PDC to know when to send the
MMC command.

Signed-off-by: Ludovic Desroches <>
Acked-by: Nicolas Ferre <>
Signed-off-by: Chris Ball <>
Signed-off-by: Ben Hutchings <>
7 years agostaging:iio:ad799x fix error_free_irq which was freeing an irq that may not have...
Hartmut Knaack [Wed, 1 Jan 2014 23:04:00 +0000 (23:04 +0000)]
staging:iio:ad799x fix error_free_irq which was freeing an irq that may not have been requested

commit 38408d056188be29a6c4e17f3703c796551bb330 upstream.

Only free an IRQ in error_free_irq, if it has been requested previously.

Signed-off-by: Hartmut Knaack <>
Acked-by: Lars-Peter Clausen <>
Signed-off-by: Jonathan Cameron <>
Signed-off-by: Ben Hutchings <>
7 years agoALSA: Enable CONFIG_ZONE_DMA for smaller PCI DMA masks
Takashi Iwai [Fri, 10 Jan 2014 13:20:42 +0000 (14:20 +0100)]
ALSA: Enable CONFIG_ZONE_DMA for smaller PCI DMA masks

commit 80ab8eae70e51d578ebbeb228e0f7a562471b8b7 upstream.

The PCI devices with DMA masks smaller than 32bit should enable
CONFIG_ZONE_DMA.  Since the recent change of page allocator, page
allocations via dma_alloc_coherent() with the limited DMA mask bits
may fail more frequently, ended up with no available buffers, when
CONFIG_ZONE_DMA isn't enabled.  With CONFIG_ZONE_DMA, the system has
much more chance to obtain such pages.

Signed-off-by: Takashi Iwai <>
Signed-off-by: Ben Hutchings <>
7 years agostaging: r8712u: Set device type to wlan
Larry Finger [Tue, 24 Dec 2013 17:22:54 +0000 (11:22 -0600)]
staging: r8712u: Set device type to wlan

commit 3a21f00a5002b14e4aab52aef59d33ed28468a13 upstream.

The latest version of NetworkManager does not recognize the device as wireless
without this change.

Signed-off-by: Larry Finger <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
7 years agoradeon/pm: Guard access to rdev->pm.power_state array
Michel Dänzer [Wed, 8 Jan 2014 02:40:20 +0000 (11:40 +0900)]
radeon/pm: Guard access to rdev->pm.power_state array

commit 370169516e736edad3b3c5aa49858058f8b55195 upstream.

It's never allocated on systems without an ATOMBIOS or COMBIOS ROM.

Should fix an oops I encountered while resetting the GPU after a lockup
on my PowerBook with an RV350.

Signed-off-by: Michel Dänzer <>
Signed-off-by: Alex Deucher <>
Signed-off-by: Ben Hutchings <>
7 years agoASoC: adau1701: Fix ADAU1701_SEROCTL_WORD_LEN_16 constant
Lars-Peter Clausen [Wed, 8 Jan 2014 10:22:25 +0000 (11:22 +0100)]
ASoC: adau1701: Fix ADAU1701_SEROCTL_WORD_LEN_16 constant

commit e20970ada3f699c113fe64b04492af083d11a7d8 upstream.

The driver defines ADAU1701_SEROCTL_WORD_LEN_16 as 0x10 while it should be b10,
so 0x2. This patch fixes it.

Reported-by: Magnus Reftel <>
Signed-off-by: Lars-Peter Clausen <>
Signed-off-by: Mark Brown <>
Signed-off-by: Ben Hutchings <>
7 years agodm space map common: make sure new space is used during extend
Joe Thornber [Tue, 7 Jan 2014 15:47:59 +0000 (15:47 +0000)]
dm space map common: make sure new space is used during extend

commit 12c91a5c2d2a8e8cc40a9552313e1e7b0a2d9ee3 upstream.

When extending a low level space map we should update nr_blocks at
the start so the new space is used for the index entries.

Otherwise extend can fail, e.g.: sm_metadata_extend call sequence
that fails:
 -> sm_ll_extend
    -> dm_tm_new_block -> dm_sm_new_block -> sm_bootstrap_new_block
    => returns -ENOSPC because smm->begin == smm->ll.nr_blocks

Signed-off-by: Joe Thornber <>
Signed-off-by: Mike Snitzer <>
Signed-off-by: Ben Hutchings <>
7 years agodm: wait until embedded kobject is released before destroying a device
Mikulas Patocka [Tue, 7 Jan 2014 04:01:22 +0000 (23:01 -0500)]
dm: wait until embedded kobject is released before destroying a device

commit be35f486108227e10fe5d96fd42fb2b344c59983 upstream.

There may be other parts of the kernel holding a reference on the dm
kobject.  We must wait until all references are dropped before
deallocating the mapped_device structure.

The dm_kobject_release method signals that all references are dropped
via completion.  But dm_kobject_release doesn't free the kobject (which
is embedded in the mapped_device structure).

This is the sequence of operations:
* when destroying a DM device, call kobject_put from dm_sysfs_exit
* wait until all users stop using the kobject, when it happens the
  release method is called
* the release method signals the completion and should return without
* the dm device removal code that waits on the completion continues
* the dm device removal code drops the dm_mod reference the device had
* the dm device removal code frees the mapped_device structure that
  contains the kobject

Using kobject this way should avoid the module unload race that was
mentioned at the beginning of this thread:

Signed-off-by: Mikulas Patocka <>
Signed-off-by: Mike Snitzer <>
Signed-off-by: Ben Hutchings <>
7 years agoserial: 8250: enable UART_BUG_NOMSR for Tegra
Stephen Warren [Tue, 7 Jan 2014 22:00:12 +0000 (15:00 -0700)]
serial: 8250: enable UART_BUG_NOMSR for Tegra

commit 3685f19e07802ec4207b52465c408f185b66490e upstream.

Tegra chips have 4 or 5 identical UART modules embedded. UARTs C..E have
their MODEM-control signals tied off to a static state. However UARTs A
and B can optionally route those signals to/from package pins, depending
on the exact pinmux configuration.

When these signals are not routed to package pins, false interrupts may
trigger either temporarily, or permanently, all while not showing up in
the IIR; it will read as NO_INT. This will eventually lead to the UART
IRQ being disabled due to unhandled interrupts. When this happens, the
kernel may print e.g.:

    irq 68: nobody cared (try booting with the "irqpoll" option)

In order to prevent this, enable UART_BUG_NOMSR. This prevents
UART_IER_MSI from being enabled, which prevents the false interrupts
from triggering.

In practice, this is not needed under any of the following conditions:

* On Tegra chips after Tegra30, since the HW bug has apparently been

* On UARTs C..E since their MODEM control signals are tied to the correct
  static state which doesn't trigger the issue.

* On UARTs A..B if the MODEM control signals are routed out to package
  pins, since they will then carry valid signals.

However, we ignore these exceptions for now, since they are only relevant
if a board actually hooks up more than a 4-wire UART, and no currently
supported board does this. If we ever support a board that does, we can
refine the algorithm that enables UART_BUG_NOMSR to take those exceptions
into account, and/or read a flag from DT/... that indicates that the
board has hooked up and pinmux'd more than a 4-wire UART.

Reported-by: Olof Johansson <> # autotester
Signed-off-by: Stephen Warren <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2:
 - Adjust filename
 - s/port->/up->port./]
Signed-off-by: Ben Hutchings <>
7 years agoUSB: ftdi_sio: added CS5 quirk for broken smartcard readers
Colin Leitner [Mon, 6 Jan 2014 20:33:54 +0000 (21:33 +0100)]
USB: ftdi_sio: added CS5 quirk for broken smartcard readers

commit c1f15196ac3b541d084dc80a8fbd8a74c6a0bd44 upstream.

Genuine FTDI chips support only CS7/8. A previous fix in commit
8704211f65a2 ("USB: ftdi_sio: fixed handling of unsupported CSIZE
setting") enforced this limitation and reported it back to userspace.

However, certain types of smartcard readers depend on specific
driver behaviour that requests 0 data bits (not 5) to change into a
different operating mode if CS5 has been set.

This patch reenables this behaviour for all FTDI devices.

Tagged to be added to stable, because it affects a lot of users of
embedded systems which rely on these readers to work properly.

Reported-by: Heinrich Siebmanns <>
Tested-by: Heinrich Siebmanns <>
Signed-off-by: Colin Leitner <>
Signed-off-by: Johan Hovold <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2: s/ddev/\&port->dev/]
Signed-off-by: Ben Hutchings <>
7 years agodrm/radeon: warn users when hw_i2c is enabled (v2)
Alex Deucher [Tue, 7 Jan 2014 15:05:02 +0000 (10:05 -0500)]
drm/radeon: warn users when hw_i2c is enabled (v2)

commit d195178297de9a91246519dbfa98952b70f9a9b6 upstream.

The hw i2c engines are disabled by default as the
current implementation is still experimental.  Print
a warning when users enable it so that it's obvious
when the option is enabled.

v2: check for non-0 rather than 1

Signed-off-by: Alex Deucher <>
Reviewed-by: Christian König <>
Signed-off-by: Ben Hutchings <>
7 years agoSELinux: Fix memory leak upon loading policy
Tetsuo Handa [Mon, 6 Jan 2014 12:28:15 +0000 (21:28 +0900)]
SELinux: Fix memory leak upon loading policy

commit 8ed814602876bec9bad2649ca17f34b499357a1c upstream.


I got below leak with linux-3.10.0-54.0.1.el7.x86_64 .

[  681.903890] kmemleak: 5538 new suspected memory leaks (see /sys/kernel/debug/kmemleak)

Below is a patch, but I don't know whether we need special handing for undoing
ebitmap_set_bit() call.
>>From fe97527a90fe95e2239dfbaa7558f0ed559c0992 Mon Sep 17 00:00:00 2001
From: Tetsuo Handa <>
Date: Mon, 6 Jan 2014 16:30:21 +0900
Subject: [PATCH] SELinux: Fix memory leak upon loading policy

Commit 2463c26d "SELinux: put name based create rules in a hashtable" did not
check return value from hashtab_insert() in filename_trans_read(). It leaks
memory if hashtab_insert() returns error.

  unreferenced object 0xffff88005c9160d0 (size 8):
    comm "systemd", pid 1, jiffies 4294688674 (age 235.265s)
    hex dump (first 8 bytes):
      57 0b 00 00 6b 6b 6b a5                          W...kkk.
      [<ffffffff816604ae>] kmemleak_alloc+0x4e/0xb0
      [<ffffffff811cba5e>] kmem_cache_alloc_trace+0x12e/0x360
      [<ffffffff812aec5d>] policydb_read+0xd1d/0xf70
      [<ffffffff812b345c>] security_load_policy+0x6c/0x500
      [<ffffffff812a623c>] sel_write_load+0xac/0x750
      [<ffffffff811eb680>] vfs_write+0xc0/0x1f0
      [<ffffffff811ec08c>] SyS_write+0x4c/0xa0
      [<ffffffff81690419>] system_call_fastpath+0x16/0x1b
      [<ffffffffffffffff>] 0xffffffffffffffff

However, we should not return EEXIST error to the caller, or the systemd will
show below message and the boot sequence freezes.

  systemd[1]: Failed to load SELinux policy. Freezing.

Signed-off-by: Tetsuo Handa <>
Acked-by: Eric Paris <>
Signed-off-by: Paul Moore <>
Signed-off-by: Ben Hutchings <>
7 years agosunrpc: Fix infinite loop in RPC state machine
Weston Andros Adamson [Tue, 17 Dec 2013 17:16:11 +0000 (12:16 -0500)]
sunrpc: Fix infinite loop in RPC state machine

commit 6ff33b7dd0228b7d7ed44791bbbc98b03fd15d9d upstream.

When a task enters call_refreshresult with status 0 from call_refresh and
!rpcauth_uptodatecred(task) it enters call_refresh again with no rate-limiting
or max number of retries.

Instead of trying forever, make use of the retry path that other errors use.

This only seems to be possible when the crrefresh callback is gss_refresh_null,
which only happens when destroying the context.

To reproduce:

1) mount with sec=krb5 (or sec=sys with krb5 negotiated for non FSID specific

2) reboot - the client will be stuck and will need to be hard rebooted

BUG: soft lockup - CPU#0 stuck for 22s! [kworker/0:2:46]
Modules linked in: rpcsec_gss_krb5 nfsv4 nfs fscache ppdev crc32c_intel aesni_intel aes_x86_64 glue_helper lrw gf128mul ablk_helper cryptd serio_raw i2c_piix4 i2c_core e1000 parport_pc parport shpchp nfsd auth_rpcgss oid_registry exportfs nfs_acl lockd sunrpc autofs4 mptspi scsi_transport_spi mptscsih mptbase ata_generic floppy
irq event stamp: 195724
hardirqs last  enabled at (195723): [<ffffffff814a925c>] restore_args+0x0/0x30
hardirqs last disabled at (195724): [<ffffffff814b0a6a>] apic_timer_interrupt+0x6a/0x80
softirqs last  enabled at (195722): [<ffffffff8103f583>] __do_softirq+0x1df/0x276
softirqs last disabled at (195717): [<ffffffff8103f852>] irq_exit+0x53/0x9a
CPU: 0 PID: 46 Comm: kworker/0:2 Not tainted 3.13.0-rc3-branch-dros_testing+ #4
Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 07/31/2013
Workqueue: rpciod rpc_async_schedule [sunrpc]
task: ffff8800799c4260 ti: ffff880079002000 task.ti: ffff880079002000
RIP: 0010:[<ffffffffa0064fd4>]  [<ffffffffa0064fd4>] __rpc_execute+0x8a/0x362 [sunrpc]
RSP: 0018:ffff880079003d18  EFLAGS: 00000246
RAX: 0000000000000005 RBX: 0000000000000007 RCX: 0000000000000007
RDX: 0000000000000007 RSI: ffff88007aecbae8 RDI: ffff8800783d8900
RBP: ffff880079003d78 R08: ffff88006e30e9f8 R09: ffffffffa005a3d7
R10: ffff88006e30e7b0 R11: ffff8800783d8900 R12: ffffffffa006675e
R13: ffff880079003ce8 R14: ffff88006e30e7b0 R15: ffff8800783d8900
FS:  0000000000000000(0000) GS:ffff88007f200000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f3072333000 CR3: 0000000001a0b000 CR4: 00000000001407f0
 ffff880079003d98 0000000000000246 0000000000000000 ffff88007a9a4830
 ffff880000000000 ffffffff81073f47 ffff88007f212b00 ffff8800799c4260
 ffff8800783d8988 ffff88007f212b00 ffffe8ffff604800 0000000000000000
Call Trace:
 [<ffffffff81073f47>] ? trace_hardirqs_on_caller+0x145/0x1a1
 [<ffffffffa00652d3>] rpc_async_schedule+0x27/0x32 [sunrpc]
 [<ffffffff81052974>] process_one_work+0x211/0x3a5
 [<ffffffff810528d5>] ? process_one_work+0x172/0x3a5
 [<ffffffff81052eeb>] worker_thread+0x134/0x202
 [<ffffffff81052db7>] ? rescuer_thread+0x280/0x280
 [<ffffffff81052db7>] ? rescuer_thread+0x280/0x280
 [<ffffffff810584a0>] kthread+0xc9/0xd1
 [<ffffffff810583d7>] ? __kthread_parkme+0x61/0x61
 [<ffffffff814afd6c>] ret_from_fork+0x7c/0xb0
 [<ffffffff810583d7>] ? __kthread_parkme+0x61/0x61
Code: e8 87 63 fd e0 c6 05 10 dd 01 00 01 48 8b 43 70 4c 8d 6b 70 45 31 e4 a8 02 0f 85 d5 02 00 00 4c 8b 7b 48 48 c7 43 48 00 00 00 00 <4c> 8b 4b 50 4d 85 ff 75 0c 4d 85 c9 4d 89 cf 0f 84 32 01 00 00

And the output of "rpcdebug -m rpc -s all":

RPC:    61 call_refresh (status 0)
RPC:    61 call_refresh (status 0)
RPC:    61 refreshing RPCSEC_GSS cred ffff88007a413cf0
RPC:    61 refreshing RPCSEC_GSS cred ffff88007a413cf0
RPC:    61 call_refreshresult (status 0)
RPC:    61 refreshing RPCSEC_GSS cred ffff88007a413cf0
RPC:    61 call_refreshresult (status 0)
RPC:    61 refreshing RPCSEC_GSS cred ffff88007a413cf0
RPC:    61 call_refresh (status 0)
RPC:    61 call_refreshresult (status 0)
RPC:    61 call_refresh (status 0)
RPC:    61 call_refresh (status 0)
RPC:    61 refreshing RPCSEC_GSS cred ffff88007a413cf0
RPC:    61 call_refreshresult (status 0)
RPC:    61 call_refresh (status 0)
RPC:    61 refreshing RPCSEC_GSS cred ffff88007a413cf0
RPC:    61 call_refresh (status 0)
RPC:    61 refreshing RPCSEC_GSS cred ffff88007a413cf0
RPC:    61 refreshing RPCSEC_GSS cred ffff88007a413cf0
RPC:    61 call_refreshresult (status 0)
RPC:    61 call_refresh (status 0)
RPC:    61 call_refresh (status 0)
RPC:    61 call_refresh (status 0)
RPC:    61 call_refresh (status 0)
RPC:    61 call_refreshresult (status 0)
RPC:    61 refreshing RPCSEC_GSS cred ffff88007a413cf0

Signed-off-by: Weston Andros Adamson <>
Signed-off-by: Trond Myklebust <>
Signed-off-by: Ben Hutchings <>
7 years agoALSA: rme9652: fix a missing comma in channel_map_9636_ds[]
Takashi Iwai [Thu, 26 Dec 2013 22:13:08 +0000 (00:13 +0200)]
ALSA: rme9652: fix a missing comma in channel_map_9636_ds[]

commit 770bd4bf2e664939a9dacd3d26ec9ff7a3933210 upstream.

The lack of comma leads to the wrong channel for an SPDIF channel.
Unfortunately this wasn't caught by compiler because it's still a
valid expression.

Reported-by: Alexander Aristov <>
Signed-off-by: Takashi Iwai <>
Signed-off-by: Ben Hutchings <>
7 years agoUSB: cypress_m8: fix ring-indicator detection and reporting
Johan Hovold [Thu, 2 Jan 2014 21:49:24 +0000 (22:49 +0100)]
USB: cypress_m8: fix ring-indicator detection and reporting

commit 440ebadeae9298d7de3d4d105342691841ec88d0 upstream.

Fix ring-indicator (RI) status-bit definition, which was defined as CTS,
effectively preventing RI-changes from being detected while reporting
false RI status.

This bug predates git.

Signed-off-by: Johan Hovold <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
7 years agoUSB: Nokia 502 is an unusual device
Mikhail Zolotaryov [Fri, 27 Dec 2013 23:56:35 +0000 (01:56 +0200)]
USB: Nokia 502 is an unusual device

commit 0e16114f2db4838251fb64f3b550996ad3585890 upstream.

The USB storage operation of Nokia Asha 502 Dual SIM smartphone running Asha
Platform 1.1.1 is unreliable in respect of data consistency (i.e. transfered
files are corrupted). A similar issue is described here:

The workaround is (MAX_SECTORS_64):
   rmmod usb_storage && modprobe usb_storage quirks=0421:06aa:m

The patch adds the tested device to the unusual list permanently.

Signed-off-by: Mikhail Zolotaryov <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
7 years agoUSB: serial: add support for iBall 3.5G connect usb modem
Rahul Bedarkar [Thu, 2 Jan 2014 15:27:56 +0000 (20:57 +0530)]
USB: serial: add support for iBall 3.5G connect usb modem

commit 7d5c1b9c7cb5ec8e52b1adc65c484a923a8ea6c3 upstream.

Add support for iBall 3.5G connect usb modem.

Bus 002 Device 006: ID 1c9e:9605 OMEGA TECHNOLOGY

T:  Bus=02 Lev=01 Prnt=01 Port=00 Cnt=01 Dev#=  6 Spd=480 MxCh= 0
D:  Ver= 2.00 Cls=00(>ifc ) Sub=00 Prot=00 MxPS=64 #Cfgs=  1
P:  Vendor=1c9e ProdID=9605 Rev=00.00
S:  Manufacturer=USB Modem
S:  Product=USB Modem
S:  SerialNumber=1234567890ABCDEF
C:  #Ifs= 5 Cfg#= 1 Atr=e0 MxPwr=500mA
I:  If#= 0 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=option
I:  If#= 1 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=option
I:  If#= 2 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=option
I:  If#= 3 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=ff Driver=option
I:  If#= 4 Alt= 0 #EPs= 2 Cls=08(stor.) Sub=06 Prot=50 Driver=usb-storage

Signed-off-by: Rahul Bedarkar <>
Suggested-by: Bjørn Mork <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
7 years agoUSB: pl2303: fix data corruption on termios updates
Johan Hovold [Sun, 29 Dec 2013 18:22:53 +0000 (19:22 +0100)]
USB: pl2303: fix data corruption on termios updates

commit 623c8263376c0b8a4b0c220232e7313d762cd0cc upstream.

Some PL2303 devices are known to lose bytes if you change serial
settings even to the same values as before. Avoid this by comparing the
encoded settings with the previsouly used ones before configuring the

The common case was fixed by commit bf5e5834bffc6 ("pl2303: Fix mode
switching regression"), but this problem was still possible to trigger,
for instance, by using the TCSETS2-interface to repeatedly request
115201 baud, which gets mapped to 115200 and thus always triggers a
settings update.

Cc: Frank Schäfer <>
Signed-off-by: Johan Hovold <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2: adjust context; use dbg() instead of dev_dbg()]
Signed-off-by: Ben Hutchings <>
7 years agomtd: mxc_nand: remove duplicated ecc_stats counting
Michael Grzeschik [Fri, 29 Nov 2013 13:14:29 +0000 (14:14 +0100)]
mtd: mxc_nand: remove duplicated ecc_stats counting

commit 0566477762f9e174e97af347ee9c865f908a5647 upstream.

The ecc_stats.corrected count variable will already be incremented in
the above framework-layer just after this callback.

Signed-off-by: Michael Grzeschik <>
Signed-off-by: Brian Norris <>
Signed-off-by: Ben Hutchings <>
7 years agoslub: Fix calculation of cpu slabs
Li Zefan [Tue, 10 Sep 2013 03:43:37 +0000 (11:43 +0800)]
slub: Fix calculation of cpu slabs

commit 8afb1474db4701d1ab80cd8251137a3260e6913e upstream.

  /sys/kernel/slab/:t-0000048 # cat cpu_slabs
  231 N0=16 N1=215
  /sys/kernel/slab/:t-0000048 # cat slabs
  145 N0=36 N1=109

See, the number of slabs is smaller than that of cpu slabs.

The bug was introduced by commit 49e2258586b423684f03c278149ab46d8f8b6700
("slub: per cpu cache for partial pages").

We should use page->pages instead of page->pobjects when calculating
the number of cpu partial slabs. This also fixes the mapping of slabs
and nodes.

As there's no variable storing the number of total/active objects in
cpu partial slabs, and we don't have user interfaces requiring those
statistics, I just add WARN_ON for those cases.

Acked-by: Christoph Lameter <>
Reviewed-by: Wanpeng Li <>
Signed-off-by: Li Zefan <>
Signed-off-by: Pekka Enberg <>
Signed-off-by: Ben Hutchings <>
7 years agortc-cmos: Add an alarm disable quirk
Borislav Petkov [Sat, 20 Jul 2013 17:00:23 +0000 (19:00 +0200)]
rtc-cmos: Add an alarm disable quirk

commit d5a1c7e3fc38d9c7d629e1e47f32f863acbdec3d upstream.

41c7f7424259f ("rtc: Disable the alarm in the hardware (v2)") added the
functionality to disable the RTC wake alarm when shutting down the box.

However, there are at least two b0rked BIOSes we know about:

where, when wakeup alarm is enabled in the BIOS, the machine reboots
automatically right after shutdown, regardless of what wakeup time is

Bisecting the issue lead to this patch so disable its functionality with
a DMI quirk only for those boxes.

Cc: Brecht Machiels <>
Cc: Thomas Gleixner <>
Cc: John Stultz <>
Cc: Rabin Vincent <>
Signed-off-by: Borislav Petkov <>
[jstultz: Changed variable name for clarity, added extra dmi entry]
Tested-by: Brecht Machiels <>
Tested-by: Borislav Petkov <>
Signed-off-by: John Stultz <>
Signed-off-by: Ben Hutchings <>
7 years agox86/efi: Fix off-by-one bug in EFI Boot Services reservation
Dave Young [Fri, 20 Dec 2013 10:02:15 +0000 (18:02 +0800)]
x86/efi: Fix off-by-one bug in EFI Boot Services reservation

commit a7f84f03f660d93574ac88835d056c0d6468aebe upstream.

Current code check boot service region with kernel text region by:
start+size >= __pa_symbol(_text)
The end of the above region should be start + size - 1 instead.

I see this problem in ovmf + Fedora 19 grub boot:
text start: 1000000 md start: 800000 md size: 800000

Signed-off-by: Dave Young <>
Acked-by: Borislav Petkov <>
Acked-by: Toshi Kani <>
Tested-by: Toshi Kani <>
Signed-off-by: Matt Fleming <>
[bwh: Backported to 3.2: s/__pa_symbol/virt_to_phys/]
Signed-off-by: Ben Hutchings <>
7 years agoserial: add support for 200 v3 series Titan card
Yegor Yefremov [Mon, 9 Dec 2013 11:11:15 +0000 (12:11 +0100)]
serial: add support for 200 v3 series Titan card

commit 48c0247d7b7bf58abb85a39021099529df365c4d upstream.

Signed-off-by: Yegor Yefremov <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2: adjust filename]
Signed-off-by: Ben Hutchings <>
7 years agoserial: add support for 400 and 800 v3 series Titan cards
Yegor Yefremov [Tue, 27 Dec 2011 14:47:37 +0000 (15:47 +0100)]
serial: add support for 400 and 800 v3 series Titan cards

commit 1e9deb118ed76b9df89d189f27a06522a03cf743 upstream.

add support for 400Hv3, 410Hv3 and 800Hv3

Signed-off-by: Yegor Yefremov <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
7 years agodib8000: make 32 bits read atomic
Mauro Carvalho Chehab [Fri, 13 Dec 2013 13:35:03 +0000 (10:35 -0300)]
dib8000: make 32 bits read atomic

commit 5ac64ba12aca3bef18e61c866583155a3bbf81c4 upstream.

As the dvb-frontend kthread can be called anytime, it can race
with some get status ioctl. So, it seems better to avoid one to
race with the other while reading a 32 bits register.
I can't see any other reason for having a mutex there at I2C, except
to provide such kind of protection, as the I2C core already has a
mutex to protect I2C transfers.

Note: instead of this approach, it could eventually remove the dib8000
specific mutex for it, and either group the 4 ops into one xfer or
to manually control the I2C mutex. The main advantage of the current
approach is that the changes are smaller and more puntual.

Signed-off-by: Mauro Carvalho Chehab <>
Signed-off-by: Mauro Carvalho Chehab <>
Acked-by: Patrick Boettcher <>
[bwh: Backported to 3.2: adjust filename]
Signed-off-by: Ben Hutchings <>
7 years agousb: option: add new zte 3g modem pids to option driver
张君 [Wed, 18 Dec 2013 07:37:17 +0000 (15:37 +0800)]
usb: option: add new zte 3g modem pids to option driver

commit 4d90b819ae4c7ea8fd5e2bb7edc68c0f334be2e4 upstream.

Signed-off-by: Jun zhang <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
7 years agobfa: Chinook quad port 16G FC HBA claim issue
Vijaya Mohan Guvva [Wed, 4 Dec 2013 13:43:58 +0000 (05:43 -0800)]
bfa: Chinook quad port 16G FC HBA claim issue

commit dcaf9aed995c2b2a49fb86bbbcfa2f92c797ab5d upstream.

Bfa driver crash is observed while pushing the firmware on to chinook
quad port card due to uninitialized bfi_image_ct2 access which gets
initialized only for CT2 ASIC based cards after request_firmware().
For quard port chinook (CT2 ASIC based), bfi_image_ct2 is not getting
initialized as there is no check for chinook PCI device ID before
request_firmware and instead bfi_image_cb is initialized as it is the
default case for card type check.

This patch includes changes to read the right firmware for quad port chinook.

Signed-off-by: Vijaya Mohan Guvva <>
Signed-off-by: James Bottomley <>
Signed-off-by: Ben Hutchings <>
7 years agostaging: vt6656: [BUG] BBvUpdatePreEDThreshold Always set sensitivity on bScanning
Malcolm Priestley [Sun, 8 Dec 2013 09:11:30 +0000 (09:11 +0000)]
staging: vt6656: [BUG] BBvUpdatePreEDThreshold Always set sensitivity on bScanning

commit 8f248dae133668bfb8e9379b4b3f0571c858b24a upstream.

byBBPreEDIndex value is initially 0, this means that from
cold BBvUpdatePreEDThreshold is never set.

This means that sensitivity may be in an ambiguous state,
failing to scan any wireless points or at least distant ones.

Signed-off-by: Malcolm Priestley <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
7 years agoparport: parport_pc: remove double PCI ID for NetMos
Sebastian Andrzej Siewior [Wed, 27 Nov 2013 16:43:43 +0000 (17:43 +0100)]
parport: parport_pc: remove double PCI ID for NetMos

commit d6a484520c5572a4170fa915109ccfc0c38f5008 upstream.

In commit 85747f ("PATCH] parport: add NetMOS 9805 support") Max added
the PCI ID for NetMOS 9805 based on a Debian bug report from 2k4 which
was at the v2.4.26 time frame. The patch made into 2.6.14.
Shortly before that patch akpm merged commit 296d3c783b ("[PATCH] Support
NetMOS based PCI cards providing serial and parallel ports") which made
into v2.6.9-rc1.
Now we have two different entries for the same PCI id.
I have here the NetMos 9805 which claims to support SPP/EPP/ECP mode.
This patch takes Max's entry for titan_1284p1 (base != -1 specifies the
ioport for ECP mode) and replaces akpm's entry for netmos_9805 which
specified -1 (=none). Both share the same PCI-ID (my card has subsystem
0x1000 / 0x0020 so it should match PCI_ANY).

While here I also drop the entry for titan_1284p2 which is the same as

Cc: Maximilian Attems <>
Signed-off-by: Sebastian Andrzej Siewior <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
7 years agoselinux: process labeled IPsec TCP SYN-ACK packets properly in selinux_ip_postroute()
Paul Moore [Mon, 9 Dec 2013 21:11:53 +0000 (16:11 -0500)]
selinux: process labeled IPsec TCP SYN-ACK packets properly in selinux_ip_postroute()

commit 5c6c26813a209e7075baf908e3ad81c1a9d389e8 upstream.

Due to difficulty in arriving at the proper security label for
TCP SYN-ACK packets in selinux_ip_postroute(), we need to check packets
while/before they are undergoing XFRM transforms instead of waiting
until afterwards so that we can determine the correct security label.

Reported-by: Janak Desai <>
Signed-off-by: Paul Moore <>
[bwh: Backported to 3.2:
 s/selinux_peerlbl_enabled()/netlbl_enabled() || selinux_xfrm_enabled()/]
Signed-off-by: Ben Hutchings <>
7 years agoNFSv4: OPEN must handle the NFS4ERR_IO return code correctly
Trond Myklebust [Wed, 4 Dec 2013 22:39:23 +0000 (17:39 -0500)]
NFSv4: OPEN must handle the NFS4ERR_IO return code correctly

commit c7848f69ec4a8c03732cde5c949bd2aa711a9f4b upstream.

decode_op_hdr() cannot distinguish between an XDR decoding error and
the perfectly valid errorcode NFS4ERR_IO. This is normally not a
problem, but for the particular case of OPEN, we need to be able
to increment the NFSv4 open sequence id when the server returns
a valid response.

Reported-by: J Bruce Fields <>
Signed-off-by: Trond Myklebust <>
Signed-off-by: Ben Hutchings <>
7 years agortlwifi: rtl8192cu: Fix some code in RF handling
Larry Finger [Mon, 18 Nov 2013 17:11:34 +0000 (11:11 -0600)]
rtlwifi: rtl8192cu: Fix some code in RF handling

commit e9b0784bb9de3152e787ee779868c626b137fb3b upstream.

Signed-off-by: Larry Finger <>
Signed-off-by: John W. Linville <>
Signed-off-by: Ben Hutchings <>
7 years agortlwifi: Set the link state
Larry Finger [Mon, 18 Nov 2013 17:11:28 +0000 (11:11 -0600)]
rtlwifi: Set the link state

commit 619ce76f8bb850b57032501a39f26aa6c6731c70 upstream.

The present code fails to set the linked state when an interface is

Signed-off-by: Larry Finger <>
Signed-off-by: John W. Linville <>
Signed-off-by: Ben Hutchings <>
7 years agoLinux 3.2.55 v3.2.55
Ben Hutchings [Sat, 15 Feb 2014 19:20:18 +0000 (19:20 +0000)]
Linux 3.2.55

7 years agosched/rt: Avoid updating RT entry timeout twice within one tick period
Ying Xue [Tue, 17 Jul 2012 07:03:43 +0000 (15:03 +0800)]
sched/rt: Avoid updating RT entry timeout twice within one tick period

commit 57d2aa00dcec67afa52478730f2b524521af14fb upstream.

The issue below was found in 2.6.34-rt rather than mainline rt
kernel, but the issue still exists upstream as well.

So please let me describe how it was noticed on 2.6.34-rt:

On this version, each softirq has its own thread, it means there
is at least one RT FIFO task per cpu. The priority of these
tasks is set to 49 by default. If user launches an RT FIFO task
with priority lower than 49 of softirq RT tasks, it's possible
there are two RT FIFO tasks enqueued one cpu runqueue at one
moment. By current strategy of balancing RT tasks, when it comes
to RT tasks, we really need to put them off to a CPU that they
can run on as soon as possible. Even if it means a bit of cache
line flushing, we want RT tasks to be run with the least latency.

When the user RT FIFO task which just launched before is
running, the sched timer tick of the current cpu happens. In this
tick period, the timeout value of the user RT task will be
updated once. Subsequently, we try to wake up one softirq RT
task on its local cpu. As the priority of current user RT task
is lower than the softirq RT task, the current task will be
preempted by the higher priority softirq RT task. Before
preemption, we check to see if current can readily move to a
different cpu. If so, we will reschedule to allow the RT push logic
to try to move current somewhere else. Whenever the woken
softirq RT task runs, it first tries to migrate the user FIFO RT
task over to a cpu that is running a task of lesser priority. If
migration is done, it will send a reschedule request to the found
cpu by IPI interrupt. Once the target cpu responds the IPI
interrupt, it will pick the migrated user RT task to preempt its
current task. When the user RT task is running on the new cpu,
the sched timer tick of the cpu fires. So it will tick the user
RT task again. This also means the RT task timeout value will be
updated again. As the migration may be done in one tick period,
it means the user RT task timeout value will be updated twice
within one tick.

If we set a limit on the amount of cpu time for the user RT task
by setrlimit(RLIMIT_RTTIME), the SIGXCPU signal should be posted
upon reaching the soft limit.

But exactly when the SIGXCPU signal should be sent depends on the
RT task timeout value. In fact the timeout mechanism of sending
the SIGXCPU signal assumes the RT task timeout is increased once
every tick.

However, currently the timeout value may be added twice per
tick. So it results in the SIGXCPU signal being sent earlier
than expected.

To solve this issue, we prevent the timeout value from increasing
twice within one tick time by remembering the jiffies value of
last updating the timeout. As long as the RT task's jiffies is
different with the global jiffies value, we allow its timeout to
be updated.

Signed-off-by: Ying Xue <>
Signed-off-by: Fan Du <>
Reviewed-by: Yong Zhang <>
Acked-by: Steven Rostedt <>
Cc: <>
Signed-off-by: Ingo Molnar <>
[ lizf: backported to 3.4: adjust context ]
Signed-off-by: Li Zefan <>
[bwh: Backported to 3.2: adjust filename]
Signed-off-by: Ben Hutchings <>
7 years agosched: Unthrottle rt runqueues in __disable_runtime()
Peter Boonstoppel [Thu, 9 Aug 2012 22:34:47 +0000 (15:34 -0700)]
sched: Unthrottle rt runqueues in __disable_runtime()

commit a4c96ae319b8047f62dedbe1eac79e321c185749 upstream.

migrate_tasks() uses _pick_next_task_rt() to get tasks from the
real-time runqueues to be migrated. When rt_rq is throttled
_pick_next_task_rt() won't return anything, in which case
migrate_tasks() can't move all threads over and gets stuck in an
infinite loop.

Instead unthrottle rt runqueues before migrating tasks.

Additionally: move unthrottle_offline_cfs_rqs() to rq_offline_fair()

Signed-off-by: Peter Boonstoppel <>
Signed-off-by: Peter Zijlstra <>
Cc: Paul Turner <>
Signed-off-by: Ingo Molnar <>
[ lizf: backported to 3.4: adjust context ]
Signed-off-by: Li Zefan <>
[bwh: Backported to 3.2:
 - Adjust filenames
 - unthrottle_offline_cfs_rqs() is already static, but defined in sched.c
   after including sched_fair.c, so add forward declaration
 - unthrottle_offline_cfs_rqs() also needs to be defined for all CONFIG_SMP
   configurations now]
Signed-off-by: Ben Hutchings <>
7 years agosched,rt: fix isolated CPUs leaving root_task_group indefinitely throttled
Mike Galbraith [Tue, 7 Aug 2012 08:02:38 +0000 (10:02 +0200)]
sched,rt: fix isolated CPUs leaving root_task_group indefinitely throttled

commit e221d028bb08b47e624c5f0a31732c642db9d19a upstream.

Root task group bandwidth replenishment must service all CPUs, regardless of
where the timer was last started, and regardless of the isolation mechanism,
lest 'Quoth the Raven, "Nevermore"' become rt scheduling policy.

Signed-off-by: Mike Galbraith <>
Signed-off-by: Peter Zijlstra <>
Signed-off-by: Thomas Gleixner <>
[bwh: Backported to 3.2: adjust filename]
Signed-off-by: Ben Hutchings <>
7 years agosched/rt: Fix SCHED_RR across cgroups
Colin Cross [Thu, 17 May 2012 04:34:23 +0000 (21:34 -0700)]
sched/rt: Fix SCHED_RR across cgroups

commit 454c79999f7eaedcdf4c15c449e43902980cbdf5 upstream.

task_tick_rt() has an optimization to only reschedule SCHED_RR tasks
if they were the only element on their rq.  However, with cgroups
a SCHED_RR task could be the only element on its per-cgroup rq but
still be competing with other SCHED_RR tasks in its parent's
cgroup.  In this case, the SCHED_RR task in the child cgroup would
never yield at the end of its timeslice.  If the child cgroup
rt_runtime_us was the same as the parent cgroup rt_runtime_us,
the task in the parent cgroup would starve completely.

Modify task_tick_rt() to check that the task is the only task on its
rq, and that the each of the scheduling entities of its ancestors
is also the only entity on its rq.

Signed-off-by: Colin Cross <>
Signed-off-by: Peter Zijlstra <>
Signed-off-by: Ingo Molnar <>
[bwh: Backported to 3.2: adjust filename]
Signed-off-by: Ben Hutchings <>
7 years agomm: hugetlbfs: fix hugetlbfs optimization
Andrea Arcangeli [Thu, 21 Nov 2013 22:32:02 +0000 (14:32 -0800)]
mm: hugetlbfs: fix hugetlbfs optimization

commit 27c73ae759774e63313c1fbfeb17ba076cea64c5 upstream.

Commit 7cb2ef56e6a8 ("mm: fix aio performance regression for database
caused by THP") can cause dereference of a dangling pointer if
split_huge_page runs during PageHuge() if there are updates to the
tail_page->private field.

Also it is repeating compound_head twice for hugetlbfs and it is running
compound_head+compound_trans_head for THP when a single one is needed in
both cases.

The new code within the PageSlab() check doesn't need to verify that the
THP page size is never bigger than the smallest hugetlbfs page size, to
avoid memory corruption.

A longstanding theoretical race condition was found while fixing the
above (see the change right after the skip_unlock label, that is
relevant for the compound_lock path too).

By re-establishing the _mapcount tail refcounting for all compound
pages, this also fixes the below problem:

  echo 0 >/sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages

  BUG: Bad page state in process bash  pfn:59a01
  page:ffffea000139b038 count:0 mapcount:10 mapping:          (null) index:0x0
  page flags: 0x1c00000000008000(tail)
  Modules linked in:
  CPU: 6 PID: 2018 Comm: bash Not tainted 3.12.0+ #25
  Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
  Call Trace:

Signed-off-by: Khalid Aziz <>
Signed-off-by: Andrea Arcangeli <>
Tested-by: Khalid Aziz <>
Cc: Pravin Shelar <>
Cc: Greg Kroah-Hartman <>
Cc: Ben Hutchings <>
Cc: Christoph Lameter <>
Cc: Johannes Weiner <>
Cc: Mel Gorman <>
Cc: Rik van Riel <>
Cc: Andi Kleen <>
Cc: Minchan Kim <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
[Khalid Aziz: Backported to 3.4]
Signed-off-by: Ben Hutchings <>
7 years agomm: fix aio performance regression for database caused by THP
Khalid Aziz [Wed, 11 Sep 2013 21:22:20 +0000 (14:22 -0700)]
mm: fix aio performance regression for database caused by THP

commit 7cb2ef56e6a8b7b368b2e883a0a47d02fed66911 upstream.

I am working with a tool that simulates oracle database I/O workload.
This tool (orion to be specific -
allocates hugetlbfs pages using shmget() with SHM_HUGETLB flag.  It then
does aio into these pages from flash disks using various common block
sizes used by database.  I am looking at performance with two of the most
common block sizes - 1M and 64K.  aio performance with these two block
sizes plunged after Transparent HugePages was introduced in the kernel.
Here are performance numbers:

pre-THP 2.6.39 3.11-rc5
1M read 8384 MB/s 5629 MB/s 6501 MB/s
64K read 7867 MB/s 4576 MB/s 4251 MB/s

I have narrowed the performance impact down to the overheads introduced by
THP in __get_page_tail() and put_compound_page() routines.  perf top shows
>40% of cycles being spent in these two routines.  Every time direct I/O
to hugetlbfs pages starts, kernel calls get_page() to grab a reference to
the pages and calls put_page() when I/O completes to put the reference
away.  THP introduced significant amount of locking overhead to get_page()
and put_page() when dealing with compound pages because hugepages can be
split underneath get_page() and put_page().  It added this overhead
irrespective of whether it is dealing with hugetlbfs pages or transparent
hugepages.  This resulted in 20%-45% drop in aio performance when using
hugetlbfs pages.

Since hugetlbfs pages can not be split, there is no reason to go through
all the locking overhead for these pages from what I can see.  I added
code to __get_page_tail() and put_compound_page() to bypass all the
locking code when working with hugetlbfs pages.  This improved performance
significantly.  Performance numbers with this patch:

pre-THP 3.11-rc5 3.11-rc5 + Patch
1M read 8384 MB/s 6501 MB/s 8371 MB/s
64K read 7867 MB/s 4251 MB/s 6510 MB/s

Performance with 64K read is still lower than what it was before THP, but
still a 53% improvement.  It does mean there is more work to be done but I
will take a 53% improvement for now.

Please take a look at the following patch and let me know if it looks

[ tweak comments]
Signed-off-by: Khalid Aziz <>
Cc: Pravin B Shelar <>
Cc: Christoph Lameter <>
Cc: Andrea Arcangeli <>
Cc: Johannes Weiner <>
Cc: Mel Gorman <>
Cc: Rik van Riel <>
Cc: Minchan Kim <>
Cc: Andi Kleen <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
7 years agoperf/x86/amd/ibs: Fix waking up from S3 for AMD family 10h
Robert Richter [Wed, 15 Jan 2014 14:57:29 +0000 (15:57 +0100)]
perf/x86/amd/ibs: Fix waking up from S3 for AMD family 10h

commit bee09ed91cacdbffdbcd3b05de8409c77ec9fcd6 upstream.

On AMD family 10h we see following error messages while waking up from
S3 for all non-boot CPUs leading to a failed IBS initialization:

 Enabling non-boot CPUs ...
 smpboot: Booting Node 0 Processor 1 APIC 0x1
 [Firmware Bug]: cpu 1, try to use APIC500 (LVT offset 0) for vector 0x400, but the register is already in use for vector 0xf9 on another cpu
 perf: IBS APIC setup failed on cpu #1
 process: Switch to broadcast mode on CPU1
 CPU1 is up
 ACPI: Waking up from system sleep state S3

Reason for this is that during suspend the LVT offset for the IBS
vector gets lost and needs to be reinialized while resuming.

The offset is read from the IBSCTL msr. On family 10h the offset needs
to be 1 as offset 0 is used for the MCE threshold interrupt, but
firmware assings it for IBS to 0 too. The kernel needs to reprogram
the vector. The msr is a readonly node msr, but a new value can be
written via pci config space access. The reinitialization is
implemented for family 10h in setup_ibs_ctl() which is forced during
IBS setup.

This patch fixes IBS setup after waking up from S3 by adding
resume/supend hooks for the boot cpu which does the offset

Marking it as stable to let distros pick up this fix.

Signed-off-by: Robert Richter <>
Signed-off-by: Peter Zijlstra <>
Cc: Linus Torvalds <>
Signed-off-by: Ingo Molnar <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
7 years agonilfs2: fix segctor bug that causes file system corruption
Andreas Rohner [Wed, 15 Jan 2014 01:56:36 +0000 (17:56 -0800)]
nilfs2: fix segctor bug that causes file system corruption

commit 70f2fe3a26248724d8a5019681a869abdaf3e89a upstream.

There is a bug in the function nilfs_segctor_collect, which results in
active data being written to a segment, that is marked as clean.  It is
possible, that this segment is selected for a later segment
construction, whereby the old data is overwritten.

The problem shows itself with the following kernel log message:

  nilfs_sufile_do_cancel_free: segment 6533 must be clean

Usually a few hours later the file system gets corrupted:

  NILFS: bad btree node (blocknr=8748107): level = 0, flags = 0x0, nchildren = 0
  NILFS error (device sdc1): nilfs_bmap_last_key: broken bmap (inode number=114660)

The issue can be reproduced with a file system that is nearly full and
with the cleaner running, while some IO intensive task is running.
Although it is quite hard to reproduce.

This is what happens:

 1. The cleaner starts the segment construction
 2. nilfs_segctor_collect is called
 3. sc_stage is on NILFS_ST_SUFILE and segments are freed
 4. sc_stage is on NILFS_ST_DAT current segment is full
 5. nilfs_segctor_extend_segments is called, which
    allocates a new segment
 6. The new segment is one of the segments freed in step 3
 7. nilfs_sufile_cancel_freev is called and produces an error message
 8. Loop around and the collection starts again
 9. sc_stage is on NILFS_ST_SUFILE and segments are freed
    including the newly allocated segment, which will contain active
    data and can be allocated at a later time
10. A few hours later another segment construction allocates the
    segment and causes file system corruption

This can be prevented by simply reordering the statements.  If
nilfs_sufile_cancel_freev is called before nilfs_segctor_extend_segments
the freed segments are marked as dirty and cannot be allocated any more.

Signed-off-by: Andreas Rohner <>
Reviewed-by: Ryusuke Konishi <>
Tested-by: Andreas Rohner <>
Signed-off-by: Ryusuke Konishi <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
7 years agohwmon: (coretemp) Fix truncated name of alarm attributes
Jean Delvare [Tue, 14 Jan 2014 14:59:55 +0000 (15:59 +0100)]
hwmon: (coretemp) Fix truncated name of alarm attributes

commit 3f9aec7610b39521c7c69d754de7265f6994c194 upstream.

When the core number exceeds 9, the size of the buffer storing the
alarm attribute name is insufficient and the attribute name is
truncated. This causes libsensors to skip these attributes as the
truncated name is not recognized.

Reported-by: Andreas Hollmann <>
Signed-off-by: Jean Delvare <>
Signed-off-by: Guenter Roeck <>
Signed-off-by: Ben Hutchings <>
7 years agomd/raid10: fix bug when raid10 recovery fails to recover a block.
NeilBrown [Sun, 5 Jan 2014 23:35:34 +0000 (10:35 +1100)]
md/raid10: fix bug when raid10 recovery fails to recover a block.

commit e8b849158508565e0cd6bc80061124afc5879160 upstream.

commit e875ecea266a543e643b19e44cf472f1412708f9
    md/raid10 record bad blocks as needed during recovery.

added code to the "cannot recover this block" path to record a bad
block rather than fail the whole recovery.
Unfortunately this new case was placed *after* r10bio was freed rather
than *before*, yet it still uses r10bio.
This is will crash with a null dereference.

So move the freeing of r10bio down where it is safe.

Fixes: e875ecea266a543e643b19e44cf472f1412708f9
Reported-by: Damian Nowak <>
Signed-off-by: NeilBrown <>
Signed-off-by: Ben Hutchings <>
7 years agomd/raid10: fix two bugs in handling of known-bad-blocks.
NeilBrown [Mon, 13 Jan 2014 23:38:09 +0000 (10:38 +1100)]
md/raid10: fix two bugs in handling of known-bad-blocks.

commit b50c259e25d9260b9108dc0c2964c26e5ecbe1c1 upstream.

If we discover a bad block when reading we split the request and
potentially read some of it from a different device.

The code path of this has two bugs in RAID10.
1/ we get a spin_lock with _irq, but unlock without _irq!!
2/ The calculation of 'sectors_handled' is wrong, as can be clearly
   seen by comparison with raid1.c

This leads to at least 2 warnings and a probable crash is a RAID10
ever had known bad blocks.

Fixes: 856e08e23762dfb92ffc68fd0a8d228f9e152160
Reported-by: Damian Nowak <>
Signed-off-by: NeilBrown <>
Signed-off-by: Ben Hutchings <>
7 years agomd/raid5: Fix possible confusion when multiple write errors occur.
NeilBrown [Mon, 6 Jan 2014 02:19:42 +0000 (13:19 +1100)]
md/raid5: Fix possible confusion when multiple write errors occur.

commit 1cc03eb93245e63b0b7a7832165efdc52e25b4e6 upstream.

commit 5d8c71f9e5fbdd95650be00294d238e27a363b5c
    md: raid5 crash during degradation

Fixed a crash in an overly simplistic way which could leave
R5_WriteError or R5_MadeGood set in the stripe cache for devices
for which it is no longer relevant.
When those devices are removed and spares added the flags are still
set and can cause incorrect behaviour.

commit 14a75d3e07c784c004b4b44b34af996b8e4ac453
    md/raid5: preferentially read from replacement device if possible.

Fixed the same bug if a more effective way, so we can now revert
the original commit.

Reported-and-tested-by: Alexander Lyakas <>
Fixes: 5d8c71f9e5fbdd95650be00294d238e27a363b5c
Signed-off-by: NeilBrown <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
7 years agoSELinux: Fix possible NULL pointer dereference in selinux_inode_permission()
Steven Rostedt [Fri, 10 Jan 2014 02:46:34 +0000 (21:46 -0500)]
SELinux: Fix possible NULL pointer dereference in selinux_inode_permission()

commit 3dc91d4338d698ce77832985f9cb183d8eeaf6be upstream.

While running stress tests on adding and deleting ftrace instances I hit
this bug:

  BUG: unable to handle kernel NULL pointer dereference at 0000000000000020
  IP: selinux_inode_permission+0x85/0x160
  PGD 63681067 PUD 7ddbe067 PMD 0
  Oops: 0000 [#1] PREEMPT
  CPU: 0 PID: 5634 Comm: ftrace-test-mki Not tainted 3.13.0-rc4-test-00033-gd2a6dde-dirty #20
  Hardware name:                  /DG965MQ, BIOS MQ96510J.86A.0372.2006.0605.1717 06/05/2006
  task: ffff880078375800 ti: ffff88007ddb0000 task.ti: ffff88007ddb0000
  RIP: 0010:[<ffffffff812d8bc5>]  [<ffffffff812d8bc5>] selinux_inode_permission+0x85/0x160
  RSP: 0018:ffff88007ddb1c48  EFLAGS: 00010246
  RAX: 0000000000000000 RBX: 0000000000800000 RCX: ffff88006dd43840
  RDX: 0000000000000001 RSI: 0000000000000081 RDI: ffff88006ee46000
  RBP: ffff88007ddb1c88 R08: 0000000000000000 R09: ffff88007ddb1c54
  R10: 6e6576652f6f6f66 R11: 0000000000000003 R12: 0000000000000000
  R13: 0000000000000081 R14: ffff88006ee46000 R15: 0000000000000000
  FS:  00007f217b5b6700(0000) GS:ffffffff81e21000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033^M
  CR2: 0000000000000020 CR3: 000000006a0fe000 CR4: 00000000000007f0
  Call Trace:
  Code: 84 a1 00 00 00 81 e3 00 20 00 00 89 d8 83 c8 02 40 f6 c6 04 0f 45 d8 40 f6 c6 08 74 71 80 cf 02 49 8b 46 38 4c 8d 4d cc 45 31 c0 <0f> b7 50 20 8b 70 1c 48 8b 41 70 89 d9 8b 78 04 e8 36 cf ff ff
  RIP  selinux_inode_permission+0x85/0x160
  CR2: 0000000000000020

Investigating, I found that the inode->i_security was NULL, and the
dereference of it caused the oops.

in selinux_inode_permission():

isec = inode->i_security;

rc = avc_has_perm_noaudit(sid, isec->sid, isec->sclass, perms, 0, &avd);

Note, the crash came from stressing the deletion and reading of debugfs
files.  I was not able to recreate this via normal files.  But I'm not
sure they are safe.  It may just be that the race window is much harder
to hit.

What seems to have happened (and what I have traced), is the file is
being opened at the same time the file or directory is being deleted.
As the dentry and inode locks are not held during the path walk, nor is
the inodes ref counts being incremented, there is nothing saving these
structures from being discarded except for an rcu_read_lock().

The rcu_read_lock() protects against freeing of the inode, but it does
not protect freeing of the inode_security_struct.  Now if the freeing of
the i_security happens with a call_rcu(), and the i_security field of
the inode is not changed (it gets freed as the inode gets freed) then
there will be no issue here.  (Linus Torvalds suggested not setting the
field to NULL such that we do not need to check if it is NULL in the
permission check).

Note, this is a hack, but it fixes the problem at hand.  A real fix is
to restructure the destroy_inode() to call all the destructor handlers
from the RCU callback.  But that is a major job to do, and requires a
lot of work.  For now, we just band-aid this bug with this fix (it
works), and work on a more maintainable solution in the future.

Signed-off-by: Steven Rostedt <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
7 years agoARM: fix "bad mode in ... handler" message for undefined instructions
Russell King [Fri, 3 Jan 2014 15:01:39 +0000 (15:01 +0000)]
ARM: fix "bad mode in ... handler" message for undefined instructions

commit 29c350bf28da333e41e30497b649fe335712a2ab upstream.

The array was missing the final entry for the undefined instruction
exception handler; this commit adds it.

Signed-off-by: Russell King <>
Signed-off-by: Ben Hutchings <>
7 years agoahci: add PCI ID for Marvell 88SE9170 SATA controller
Simon Guinot [Mon, 23 Dec 2013 12:24:35 +0000 (13:24 +0100)]
ahci: add PCI ID for Marvell 88SE9170 SATA controller

commit e098f5cbe9d410e7878b50f524dce36cc83ec40e upstream.

This patch adds support for the PCI ID provided by the Marvell 88SE9170
SATA controller.

Signed-off-by: Simon Guinot <>
Signed-off-by: Tejun Heo <>
Signed-off-by: Ben Hutchings <>
7 years agopci: Add PCI_DEVICE_SUB() macro
Ben Hutchings [Sun, 9 Feb 2014 23:34:11 +0000 (23:34 +0000)]
pci: Add PCI_DEVICE_SUB() macro

This was added as part of commit 3d567e0e291c ('tg3: Set 10_100_ONLY
flag for additional 10/100 Mbps devices') upstream and is needed by
the following patch to ahci.

Signed-off-by: Ben Hutchings <>
7 years agoahci: add an observed PCI ID for Marvell 88se9172 SATA controller
George Spelvin [Wed, 29 May 2013 01:20:35 +0000 (10:20 +0900)]
ahci: add an observed PCI ID for Marvell 88se9172 SATA controller

commit fcce9a35f8faaa1f52236c554ef1b15d99a7537e upstream.

A third possible PCI ID, as personally observed, and found in the
pci.ids list.

Signed-off-by: George Spelvin <>
Signed-off-by: Tejun Heo <>
Signed-off-by: Ben Hutchings <>
7 years agoahci: Use PCI_VENDOR_ID_MARVELL_EXT for 0x1b4b
Myron Stowe [Mon, 8 Apr 2013 17:32:49 +0000 (11:32 -0600)]
ahci: Use PCI_VENDOR_ID_MARVELL_EXT for 0x1b4b

commit 69fd3157363935b1e052bd76b8f8ec65e494306e upstream.

With the 0x1b4b vendor ID #define in place, convert hard-coded ID

Signed-off-by: Myron Stowe <>
Signed-off-by: Bjorn Helgaas <>
Acked-by: Jeff Garzik <>
Signed-off-by: Ben Hutchings <>
7 years agopowerpc: Fix bad stack check in exception entry
Michael Neuling [Mon, 16 Dec 2013 04:12:43 +0000 (15:12 +1100)]
powerpc: Fix bad stack check in exception entry

commit 90ff5d688e61f49f23545ffab6228bd7e87e6dc7 upstream.

In EXCEPTION_PROLOG_COMMON() we check to see if the stack pointer (r1)
is valid when coming from the kernel.  If it's not valid, we die but
with a nice oops message.

Currently we allocate a stack frame (subtract INT_FRAME_SIZE) before we
check to see if the stack pointer is negative.  Unfortunately, this
won't detect a bad stack where r1 is less than INT_FRAME_SIZE.

This patch fixes the check to compare the modified r1 with
-INT_FRAME_SIZE.  With this, bad kernel stack pointers (including NULL
pointers) are correctly detected again.

Kudos to Paulus for finding this.

Signed-off-by: Michael Neuling <>
Signed-off-by: Benjamin Herrenschmidt <>
Signed-off-by: Ben Hutchings <>
7 years agoARM: fix footbridge clockevent device
Russell King [Sun, 29 Dec 2013 12:39:50 +0000 (12:39 +0000)]
ARM: fix footbridge clockevent device

commit 4ff859fe1dc0da0f87bbdfff78f527898878fa4a upstream.

The clockevents code was being told that the footbridge clock event
device ticks at 16x the rate which it actually does.  This leads to
timekeeping problems since it allows the clocksource to wrap before
the kernel notices.  Fix this by using the correct clock.

Fixes: 4e8d76373c9fd ("ARM: footbridge: convert to clockevents/clocksource")
Signed-off-by: Russell King <>
[bwh: Backported to 3.2: fold in the relevant parts of commit 838a2ae80a6a
 ('ARM: use clockevents_config_and_register() where possible')]
Signed-off-by: Ben Hutchings <>
7 years agoselinux: selinux_setprocattr()->ptrace_parent() needs rcu_read_lock()
Oleg Nesterov [Mon, 23 Dec 2013 22:45:01 +0000 (17:45 -0500)]
selinux: selinux_setprocattr()->ptrace_parent() needs rcu_read_lock()

commit c0c1439541f5305b57a83d599af32b74182933fe upstream.

selinux_setprocattr() does ptrace_parent(p) under task_lock(p),
but task_struct->alloc_lock doesn't pin ->parent or ->ptrace,
this looks confusing and triggers the "suspicious RCU usage"
warning because ptrace_parent() does rcu_dereference_check().

And in theory this is wrong, spin_lock()->preempt_disable()
doesn't necessarily imply rcu_read_lock() we need to access
the ->parent.

Reported-by: Evan McNabb <>
Signed-off-by: Oleg Nesterov <>
Signed-off-by: Paul Moore <>
Signed-off-by: Ben Hutchings <>
7 years agoselinux: fix broken peer recv check
Chad Hanson [Mon, 23 Dec 2013 22:45:01 +0000 (17:45 -0500)]
selinux: fix broken peer recv check

commit 46d01d63221c3508421dd72ff9c879f61053cffc upstream.

Fix a broken networking check. Return an error if peer recv fails.  If
secmark is active and the packet recv succeeds the peer recv error is

Signed-off-by: Chad Hanson <>
Signed-off-by: Paul Moore <>
Signed-off-by: Ben Hutchings <>
7 years agodrm/radeon: 0x9649 is SUMO2 not SUMO
Alex Deucher [Mon, 23 Dec 2013 14:31:58 +0000 (09:31 -0500)]
drm/radeon: 0x9649 is SUMO2 not SUMO

commit d00adcc8ae9e22eca9d8af5f66c59ad9a74c90ec upstream.

Fixes rendering corruption due to incorrect
gfx configuration.


Signed-off-by: Alex Deucher <>
Signed-off-by: Ben Hutchings <>
7 years agoext4: add explicit casts when masking cluster sizes
Theodore Ts'o [Fri, 20 Dec 2013 14:29:35 +0000 (09:29 -0500)]
ext4: add explicit casts when masking cluster sizes

commit f5a44db5d2d677dfbf12deee461f85e9ec633961 upstream.

The missing casts can cause the high 64-bits of the physical blocks to
be lost.  Set up new macros which allows us to make sure the right
thing happen, even if at some point we end up supporting larger
logical block numbers.

Thanks to the Emese Revfy and the PaX security team for reporting this

Reported-by: PaX Team <>
Reported-by: Emese Revfy <>
Signed-off-by: "Theodore Ts'o" <>
[bwh: Backported to 3.2:
 - Adjust context
 - Drop inapplicable change to ext4_ext_rm_leaf()]
Signed-off-by: Ben Hutchings <>
7 years agodm9601: work around tx fifo sync issue on dm962x
Peter Korsgaard [Mon, 16 Dec 2013 10:35:35 +0000 (11:35 +0100)]
dm9601: work around tx fifo sync issue on dm962x

commit 4263c86dca5198da6bd3ad826d0b2304fbe25776 upstream.

Certain dm962x revisions contain an bug, where if a USB bulk transfer retry
(E.G. if bulk crc mismatch) happens right after a transfer with odd or
maxpacket length, the internal tx hardware fifo gets out of sync causing
the interface to stop working.

Work around it by adding up to 3 bytes of padding to ensure this situation
cannot trigger.

This workaround also means we never pass multiple-of-maxpacket size skb's
to usbnet, so the length adjustment to handle usbnet's padding of those can
be removed.

Reported-by: Joseph Chang <>
Signed-off-by: Peter Korsgaard <>
Signed-off-by: David S. Miller <>
Signed-off-by: Ben Hutchings <>
7 years agodm9601: fix reception of full size ethernet frames on dm9620/dm9621a
Peter Korsgaard [Mon, 16 Dec 2013 10:35:33 +0000 (11:35 +0100)]
dm9601: fix reception of full size ethernet frames on dm9620/dm9621a

commit 407900cfb54bdb2cfa228010b6697305f66b2948 upstream.

dm9620/dm9621a require room for 4 byte padding even in dm9601 (3 byte
header) mode.

Signed-off-by: Peter Korsgaard <>
Signed-off-by: David S. Miller <>
Signed-off-by: Ben Hutchings <>
7 years agonet_dma: mark broken
Dan Williams [Tue, 17 Dec 2013 18:09:32 +0000 (10:09 -0800)]
net_dma: mark broken

commit 77873803363c9e831fc1d1e6895c084279090c22 upstream.

net_dma can cause data to be copied to a stale mapping if a
copy-on-write fault occurs during dma.  The application sees missing

The following trace is triggered by modifying the kernel to WARN if it
ever triggers copy-on-write on a page that is undergoing dma:

 WARNING: CPU: 24 PID: 2529 at lib/dma-debug.c:485 debug_dma_assert_idle+0xd2/0x120()
 ioatdma 0000:00:04.0: DMA-API: cpu touching an active dma mapped page [pfn=0x16bcd9]
 Modules linked in: iTCO_wdt iTCO_vendor_support ioatdma lpc_ich pcspkr dca
 CPU: 24 PID: 2529 Comm: linbug Tainted: G        W    3.13.0-rc1+ #353
  00000000000001e5 ffff88016f45f688 ffffffff81751041 ffff88017ab0ef70
  ffff88016f45f6d8 ffff88016f45f6c8 ffffffff8104ed9c ffffffff810f3646
  ffff8801768f4840 0000000000000282 ffff88016f6cca10 00007fa2bb699349
 Call Trace:
  [<ffffffff81751041>] dump_stack+0x46/0x58
  [<ffffffff8104ed9c>] warn_slowpath_common+0x8c/0xc0
  [<ffffffff810f3646>] ? ftrace_pid_func+0x26/0x30
  [<ffffffff8104ee86>] warn_slowpath_fmt+0x46/0x50
  [<ffffffff8139c062>] debug_dma_assert_idle+0xd2/0x120
  [<ffffffff81154a40>] do_wp_page+0xd0/0x790
  [<ffffffff811582ac>] handle_mm_fault+0x51c/0xde0
  [<ffffffff813830b9>] ? copy_user_enhanced_fast_string+0x9/0x20
  [<ffffffff8175fc2c>] __do_page_fault+0x19c/0x530
  [<ffffffff8175c196>] ? _raw_spin_lock_bh+0x16/0x40
  [<ffffffff810f3539>] ? trace_clock_local+0x9/0x10
  [<ffffffff810fa1f4>] ? rb_reserve_next_event+0x64/0x310
  [<ffffffffa0014c00>] ? ioat2_dma_prep_memcpy_lock+0x60/0x130 [ioatdma]
  [<ffffffff8175ffce>] do_page_fault+0xe/0x10
  [<ffffffff8175c862>] page_fault+0x22/0x30
  [<ffffffff81643991>] ? __kfree_skb+0x51/0xd0
  [<ffffffff813830b9>] ? copy_user_enhanced_fast_string+0x9/0x20
  [<ffffffff81388ea2>] ? memcpy_toiovec+0x52/0xa0
  [<ffffffff8164770f>] skb_copy_datagram_iovec+0x5f/0x2a0
  [<ffffffff8169d0f4>] tcp_rcv_established+0x674/0x7f0
  [<ffffffff816a68c5>] tcp_v4_do_rcv+0x2e5/0x4a0
 ---[ end trace e30e3b01191b7617 ]---
 Mapped at:
  [<ffffffff8139c169>] debug_dma_map_page+0xb9/0x160
  [<ffffffff8142bf47>] dma_async_memcpy_pg_to_pg+0x127/0x210
  [<ffffffff8142cce9>] dma_memcpy_pg_to_iovec+0x119/0x1f0
  [<ffffffff81669d3c>] dma_skb_copy_datagram_iovec+0x11c/0x2b0
  [<ffffffff8169d1ca>] tcp_rcv_established+0x74a/0x7f0:

...the problem is that the receive path falls back to cpu-copy in
several locations and this trace is just one of the areas.  A few
options were considered to fix this:

1/ sync all dma whenever a cpu copy branch is taken

2/ modify the page fault handler to hold off while dma is in-flight

Option 1 adds yet more cpu overhead to an "offload" that struggles to compete
with cpu-copy.  Option 2 adds checks for behavior that is already documented as
broken when using get_user_pages().  At a minimum a debug mode is warranted to
catch and flag these violations of the dma-api vs get_user_pages().

Thanks to David for his reproducer.

Cc: Dave Jiang <>
Cc: Vinod Koul <>
Cc: Alexander Duyck <>
Reported-by: David Whipple <>
Acked-by: David S. Miller <>
Signed-off-by: Dan Williams <>
Signed-off-by: Ben Hutchings <>
7 years agoASoC: wm8904: fix DSP mode B configuration
Bo Shen [Wed, 18 Dec 2013 03:26:23 +0000 (11:26 +0800)]
ASoC: wm8904: fix DSP mode B configuration

commit f0199bc5e3a3ec13f9bc938556517ec430b36437 upstream.

When wm8904 work in DSP mode B, we still need to configure it to
work in DSP mode. Or else, it will work in Right Justified mode.

Signed-off-by: Bo Shen <>
Acked-by: Charles Keepax <>
Signed-off-by: Mark Brown <>
Signed-off-by: Ben Hutchings <>
7 years agocpupower: Fix segfault due to incorrect getopt_long arugments
Josh Boyer [Fri, 11 Oct 2013 12:45:51 +0000 (08:45 -0400)]
cpupower: Fix segfault due to incorrect getopt_long arugments

commit f447ef4a56dee4b68a91460bcdfe06b5011085f2 upstream.

If a user calls 'cpupower set --perf-bias 15', the process will end with
a SIGSEGV in libc because cpupower-set passes a NULL optarg to the atoi
call.  This is because the getopt_long structure currently has all of
the options as having an optional_argument when they really have a
required argument.  We change the structure to use required_argument to
match the short options and it resolves the issue.

This fixes

Signed-off-by: Josh Boyer <>
Cc: Dominik Brodowski <>
Cc: Thomas Renninger <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
7 years agoath9k: Fix interrupt handling for the AR9002 family
Sujith Manoharan [Mon, 16 Dec 2013 01:34:59 +0000 (07:04 +0530)]
ath9k: Fix interrupt handling for the AR9002 family

commit 73f0b56a1ff64e7fb6c3a62088804bab93bcedc2 upstream.

This patch adds a driver workaround for a HW issue.

A race condition in the HW results in missing interrupts,
which can be avoided by a read/write with the ISR register.
All chips in the AR9002 series are affected by this bug - AR9003
and above do not have this problem.

Cc: Felix Fietkau <>
Signed-off-by: Sujith Manoharan <>
Signed-off-by: John W. Linville <>
Signed-off-by: Ben Hutchings <>
7 years agortlwifi: pci: Fix oops on driver unload
Larry Finger [Wed, 11 Dec 2013 23:13:10 +0000 (17:13 -0600)]
rtlwifi: pci: Fix oops on driver unload

commit 9278db6279e28d4d433bc8a848e10b4ece8793ed upstream.

On Fedora systems, unloading rtl8192ce causes an oops. This patch fixes the
problem reported at

Signed-off-by: Larry Finger <>
Signed-off-by: John W. Linville <>
Signed-off-by: Ben Hutchings <>